Single-launch Mars sample return via hybrid chemical–electric propulsion and autonomous quadruped robotic retrieval

Abstract

The Mars Sample Return (MSR) mission is a major but prohibitively expensive goal of planetary exploration. This article presents a conceptual single-launch integrated mission architecture that consolidates entry, descent, and landing (EDL), surface sample retrieval, Mars ascent, and Earth-return functions (traditionally distributed across three separate spacecraft) into one capsule-class vehicle delivered by one heavy-lift expendable launcher. Here, “single-launch” applies only to a single launch from Earth to the Moon. The subsequent ascent from the Martian surface in the Mars Ascent–Earth Return Vehicle (MA-ERV) is an internal phase of the mission and not a separate launch campaign. The architecture consists of a three-stage Mars Ascent–Earth Return Vehicle (MA-ERV) composed of liquid-bipropellant and solid-propellant chemical stages coupled with a Hall-effect electric propulsion stage for low-thrust spiral escape. Patch-conic trajectory analysis for the 2029 primary and 2031 backup launch windows and pork chop plot evaluation and interplanetary return trajectory trade study (ballistic coast vs. continued electric propulsion thrusting) with gravity loss and drag corrections results in a total mission ΔV of 9.14 km/s. According to Edelbaum’s low-thrust approximation, with validation against numerical propagation of orbits (2%–5% discrepancy), within the wider literature of electric propulsion trajectory optimization, eclipse, J₂ perturbation, and thrust degradation sensitivity analysis, prediction of Mars escape occurs in approximately 270 days (85% thrust duty cycle) using 49 kg Xe at 5.0 kW input power. A three-degrees-of-freedom Monte Carlo EDL simulation (10,000 cases), complemented by qualitative six-degrees-of-freedom sensitivity analysis and parametric thermal protection system (TPS) mass sizing, results in a 99th-percentile landing ellipse of 8.7 km × 4.2 km with peak heat flux within PICA-class thermal protection limits. A physics-based four-legged robot energy model with penalties for Martian regolith interaction, cross-validated against terrestrial data from the Spot platform, predicts 6.2-h long sorties with a 4.6 km operational radius using in-situ resource utilization (ISRU) produced methane–oxygen fuel cells in 40% system efficiency. Mass estimation using AIAA S-120A-compliant, subsystem-level mass growth allowances and lognormal probabilistic analysis results in a total launch mass of approximately 18,200 kg using current best estimates, which is some 8% above the Falcon Heavy expendable capacity of 16,800 kg. This outcome highlights mass budget closure as the primary technical issue and encourages the development of higher capability launch vehicles (such as Falcon Heavy with performance improvements, Vulcan Centaur Heavy, or New Glenn) or subsystem mass reduction. Sensitivity and risk analyses confirm the feasibility of the conceptual architecture under nominal conditions but show that subsystem-level mass growth characteristic of technology readiness level (TRL) 3–4 technologies makes closure on the baseline Falcon Heavy launcher with no mitigation impossible. A preliminary planetary protection compliance assessment shows the Earth Return Vehicle containment architecture as a critical design driver that requires 420 kg allocation. The proposed architecture has an estimated mission cost of $3.0–6.0 billion, which is a potential cost reduction compared with similar remaining elements of the current campaign by the National Aeronautics and Space Administration/European Space Age (NASA/ESA) Mars Surveyor (MSR) campaign (estimated to be $5–8 billion without accounting for the already-operational Perseverance rover). This study identifies a mass-critical but physically plausible pathway for near-term Mars sample return at the conceptual design level while acknowledging that significant technology maturation and high fidelity design validation and advance to Phase A requires a possible launcher upgrade.

1 Introduction

1.1 Background and motivation

The return of geological samples from Mars to terrestrial laboratories has been identified as a highest-priority objective by the planetary science community for more than 3 decades (National Research Council, 2011; iMARS Working Group, 2008; McLennan et al., 2012). Unlike in situ analyses constrained by instrument mass and power limitations, laboratory examination of Martian samples would enable isotopic dating at sub-million-year precision, microscale mineralogical characterization below 100 nm resolution, and definitive biosignature detection using instruments orders of magnitude more capable than any that could be landed on Mars (Beaty et al., 2019; Carrier et al., 2020).

NASA and ESA jointly developed a baseline MSR architecture comprising three separate spacecraft: a Sample Retrieval Lander carrying a Mars Ascent Vehicle and fetch helicopters, an Earth Return Orbiter developed by ESA, and the Perseverance rover already caching samples on Mars (Muirhead et al., 2020a). The independent review board chaired by Figueroa estimated the total mission cost at $8–11 billion with a return no earlier than the mid-2030s (Figueroa et al., 2023). This cost escalation led to program restructuring in 2024–2025 (U.S. Congress, 2024).

Simultaneously, China’s Tianwen-3 mission targets the 2028–2030 window for an independent MSR attempt using a two-launch architecture (Zou et al., 2021; Wan et al., 2020). The scientific imperative for MSR remains unchanged, motivating investigation of architectures that reduce cost and schedule while maintaining scientific return.

1.2 Literature review

Mars sample return concepts have been studied since the 1970s. Early NASA concepts envisioned multi-launch architectures with sample masses exceeding 5 kg (Hoffman and Kaplan, 1997). The Mars Direct architecture introduced by Zubrin et al. (1991) and Zubrin (1996) demonstrated that combining ascent and Earth-return functions in a single in-situ resource utilization (ISRU)-fueled vehicle could yield dramatic cost reductions. Mattingly and May (2011) analyzed single-launch MSR concepts using heavy-lift vehicles.

For Mars EDL, heritage from the Mars Science Laboratory sky-crane system (Way et al., 2014) and the canceled Red Dragon propulsive-landing concept (Karcz et al., 2012) provides relevant design data. Braun and Manning (2007) established parametric relationships governing Mars EDL for payloads in the 1,000–10,000 kg class, demonstrating that propulsive deceleration becomes necessary above approximately 2,000 kg landed mass. Edquist et al. (2009) characterized supersonic retropropulsion aerodynamics for Mars landers.

Low-thrust trajectory optimization for Mars missions has a vast literature. The classical approximation for the low-thrust orbit transfer ΔV that is still foundational for preliminary electric propulsion mission design was derived by Edelbaum (1961). Betts (1998) surveyed numerical methods that apply to low-thrust trajectory optimization, including direct collocation and shooting methods. Shape-based algorithms for gravity assist trajectories were developed by Petropoulos and Longuski (2004). More recently, Whiffen and Sims (2002) presented static/dynamic control methods for low-thrust escape and capture trajectory optimization. Lantoine and Russell (2012) introduced hybrid differential dynamic programming methods that allow efficient solution of multi-phase low-thrust problems. Yam et al. (2010) generalized the low-thrust optimization to trajectories using gravity assists, relevant for alternative return architectures. Mengali and Quarta (2008) built optimal low-thrust trajectories for Earth–Mars transfers specifically to serve as analytical benchmarks for the return phase of MSR missions. Conway (2010) provides an extensive treatment of the spacecraft trajectory optimization methods. Taheri and Junkins (2018) used generic smoothing methods to enhance the convergence of low-thrust solvers for bang-off-bang thrust profiles. Onozaki et al. (2017) presented the latest trajectory design approaches for Mars missions by electric propulsion, taking advantage of the operational experience of Hayabusa2 by JAXA. Operationally, the Dawn mission (Rayman et al., 2006) demonstrated multi-year EP spiral trajectories in the inner solar system, providing critical heritage data for long-duration Hall thruster operations of the type required by the present architecture. The present study employs Edelbaum’s approximation for preliminary sizing while realizing that the more sophisticated optimization methods mentioned above are needed for refinement of the Phase A trajectory design.

Several groups have investigated quadruped robots for planetary exploration. ETH Zurich’s ANYmal platform has been studied for planetary surface operations (Hutter et al., 2017), while Boston Dynamics’ Spot robot has demonstrated relevant terrain traversal capabilities (Boston Dynamics, 2024). Arm et al. (2019) conducted analog field tests of legged robots for geological sampling. However, no prior study has performed detailed energy modeling of quadruped robots for MSR sample retrieval operations.

While the individual technologies listed above have been studied in isolation, no prior work has performed an integrated, mass-consistent, single-launch MSR architecture study that simultaneously addresses (i) hybrid chemical–electric staged ascent with validated low-thrust trajectory analysis, (ii) Monte Carlo EDL dispersion for capsule-class propulsive landing, (iii) physics-based energy modeling of quadruped robots for sample retrieval with ISRU fuel cell extension, and (iv) system-level mass and cost closure on a commercially available launcher. The present study fills this gap by coupling all four elements within a unified mission framework, enabling cross-subsystem trade studies not possible in decoupled analyses.

Planetary protection requirements for Mars sample return missions have been well analyzed. The COSPAR Planetary Protection Policy (COSPAR, 2021) classifies MSR as Category V Restricted Earth Return, requiring break-the-chain containment to avoid uncontrolled release of Martian material to the Earth’s biosphere. Containment verification needs have been analyzed by Summons et al. (2014), and the mass and complexity implications of a triple-containment architecture for Earth-return vehicles have been analyzed by Muirhead et al. (2020b). These requirements impose severe mass and design requirements that must be modeled into any credible MSR architecture study.

Surface communications for Mars robotic operations have been examined in the context of multi-robot coordination. Fong et al. (2013) showed surface communication relay ideas for planetary exploration, and Huntsberger et al. (2003) studied autonomous multi-robot coordination schemes applicable to sample retrieval missions.

ISRU systems for Mars have been analyzed extensively (Rapp, 2018; Sanders and Larson, 2013), with the MOXIE experiment providing the first in situ demonstration of oxygen production from Martian CO₂ (Hoffman et al., 2022). Sabatier-based methane production systems have been tested at laboratory scale (Zubrin et al., 2013; Muscatello and Santiago-Maldonado, 2012). The feasibility of using wind energy on Mars has been assessed by Holstein-Rathlou et al. (2010) and Lichter and Viterna (1999). However, no prior study has performed detailed energy modeling of quadruped robots for MSR sample retrieval operations.

1.3 Contributions

1.3.1 Scope and limitations

This research is at the conceptual design level (pre-Phase A) using parametric models, analytical approximations, and simplified simulations. The analyses are meant to create physical plausibility and identify critical design drivers and are not meant to provide engineering-level design verification. Key simplifications, such as 3-DOF EDL modeling, analytical structural mass estimation, and parametric cost modeling, are explicitly recognized, and the higher-fidelity analyses are identified as important future efforts before the architecture can move to Phase A study.

1.4 Article organization

The remainder of this article is structured as follows: Section 2 presents the mission architecture. Section 3 details trajectory analysis, including low-thrust optimization. Section 4 presents the EDL simulation. Section 5 covers the MA-ERV design. Section 6 describes the quadruped robot system. Section 7 addresses ISRU and energy systems. Section 8 presents mass and cost estimation. Section 9 discusses feasibility and risks. Section 10 concludes the article.

2 Mission architecture overview

2.1 Design philosophy

The architecture is controlled by four principles: (1) minimize Earth launches to one, (2) maximize heritage hardware utilization, (3) consolidate traditionally separate spacecraft functions, and (4) maintain redundancy in the sample retrieval chain. One Earth launch takes the complete integrated spacecraft to Mars; the ascent of the MA-ERV from the Martian surface is an internal mission operation, not a separate launch campaign. This is the most significant architectural difference from the baseline proposed by the space agency of the United States and the European Space Agency, which requires two or more separate launches from the Earth for the Sample Retrieval Lander, the Earth Return Orbiter, and the Perseverance rover (Muirhead et al., 2020a), and from the Chinese space agency, which uses two Earth launches for Tianwen-3 (Zou et al., 2021). The vehicle platform is specified as a capsule-class entry vehicle with characteristics similar to existing crew capsules (3.7 m base diameter, 4,200 kg dry mass class), making the architecture applicable to multiple vehicle platforms internationally.

2.2 Architecture description

The mission employs a single Falcon Heavy expendable launch to deliver an integrated spacecraft to Mars via a Type I transfer. The capsule serves as a cruise stage, aeroshell, and surface platform. Within it, the MA-ERV occupies a central structural tube (1.3 m diameter), while three quadruped robots are stowed in the annular space surrounding the MA-ERV base.

The MA-ERV comprises three stages: a liquid-bipropellant [nitrogen tetroxide/monomethylhydrazine (NTO/MMH)] first stage, a solid-propellant second stage, and a Hall-effect electric propulsion third stage. The first two stages provide impulsive ΔV from the Mars surface to low Mars orbit; the third stage executes a low-thrust spiral escape and interplanetary transfer.

Table 1 summarizes the mission elements with heritage basis and technology readiness levels.

Table 1 identifies all major spacecraft elements, their functional roles, the heritage systems from which they derive, and the assessed technology readiness level (TRL). TRL assessments follow NASA NPR 7123.1C definitions (NASA, 2016). Elements with TRL <5 require focused technology development programs prior to mission confirmation.

2.3 Concept of operations

The mission proceeds through seven phases spanning approximately 40 months.

2.3.1 Phase 1: launch and transit (months 0–9)

The launch vehicle injects the spacecraft into a Mars transfer orbit. Solar panels provide cruise power; Draco-class thrusters perform trajectory correction maneuvers.

2.3.2 Phase 2: Mars EDL (month 9)

Direct atmospheric entry from a hyperbolic approach. Aerodynamic deceleration, supersonic parachute deployment, heat shield jettison, and propulsive terminal descent achieve landing within 10 km of Perseverance’s location in Jezero Crater.

2.3.3 Phase 3: surface operations (months 9–20)

Quadruped robots deploy, retrieve cached sample tubes, and transport them to the capsule for loading into the ERV. The ISRU system produces methane and oxygen to refuel the fuel cells.

2.3.4 Phase 4: MA-ERV Mars surface ascent (month ∼20)

Timed for the Earth-return window, the MA-ERV fires up and blasts through the capsule’s middle tube. Functionally, this ascent from the Martian surface is the same as the Mars Ascent Vehicle in the example of a Mars lander, which is the baseline of the current joint Mars exploration plan by the United States and Europe (NASA/ESA), but here, the ascent capability is integrated into the same landed platform, instead of being provided by a separate spacecraft.

2.3.5 Phase 5: Mars escape (months 20–26)

After chemical stages reach low Mars orbit (LMO), the Hall thruster executes a spiral escape over ∼5 months.

2.3.6 Phase 6: interplanetary transfer (months 26–38)

After departure from the Mars sphere of influence, the spacecraft is on a heliocentric Type I or Type II return orb. The baseline design assumes the use of a ballistic coast with onboard chemical reaction control system (RCS) midcourse corrections of 0.1–0.2 km/s. An alternative configuration extending the operation of a Hall thruster into the heliocentric phase could shave 2–3 months off the transit time at a cost of 8–15 kg additional xenon.

2.3.7 Phase 7: earth entry and recovery (month ∼40)

The ERV performs skip-entry and parachute recovery.

Figure 1 shows the mission phase sequence diagram. The main mission timeline flows top-to-bottom through seven phases. Surface operations (Phase 3) include an iterative loop of robot deployment, sample retrieval, ISRU fuel production, and robot refueling that continues until all accessible sample tubes are collected.

FIGURE 1

3 Trajectory analysis

3.1 Interplanetary transfer

This Earth-to-Mars transfer is Type I. A pork chop plot analysis was conducted for both primary and backup launch windows for 2029 and 2031, respectively, using patched-conic methodology according to (Battin 1999; Curtis 2020).

For the 2029 window, the optimum departure date is within the range of May–July 2029, with the minimum C₃ of ∼8.5 km²/s² occurring around 15 June 2029, with a corresponding Mars arrival of 2.65 km/s for a 7.5-month transfer. The 21-day launch window (C₃ < 10 km²/s²) produces arrival values of 2.65–3.5 km/s with transit durations of 7–9 months. A somewhat higher minimum C₃ of approximately 10.2 km²/s² is possible due to synodic-cycle phasing for the 2031 backup window, with transit time lengths of 8–9 months and arrival of 2.9–3.3 km/s.

The sensitivity of the payload capacity of Falcon Heavy to departure C₃ is an important constraint. At the optimum C₃ of 8.5 km²/s², the expendable Falcon Heavy delivers approximately 16,800 kg to Mars transfer orbit (SpaceX, 2026). For every 1 km²/s² increase in C₃, available payload is reduced by approximately 500–700 kg, which means that the 21-day launch window limits maximum injected mass to approximately 15,800–16,800 kg. This coupling between the launch date flexibility and available mass reinforces the mass-critical nature of the architecture identified in Section 8.

The outbound, interplanetary transfer is ballistic (unpowered coast) after trans-Mars injection. Midcourse correction maneuvers (approximately 20–50 m/s) using the capsule’s Draco-class reaction control thrusters are included. Navigation accuracy that can be achieved with the use of X-band Doppler tracking (Folkner et al., 1997) provides an accuracy sufficient to target Mars approach to within the EDL entry corridor defined in Section 4.

3.2 Mars ascent ΔV budget

The total ΔV for the MA-ERV is derived with explicit accounting for all loss mechanisms. Table 2 provides an itemized breakdown of the velocity increment required for each Mars departure phase. Gravity losses are estimated using the parametric correlation from Wertz et al. (2011) for a vehicle with a thrust-to-weight ratio of 1.5–2.5 on Mars. Atmospheric drag is integrated over the ascent trajectory through the first 50 km using the Mars global reference atmospheric model (Mars-GRAM) (Justus and Johnson, 2001) with surface density ρ₀ ≈ 0.017 kg/m³ at Jezero Crater elevation [−2.5 km MOLA (Smith et al., 2001)]. The ideal ascent ΔV of 3.65 km/s is computed from Hohmann transfer mechanics: for the surface burn, plus for circularization at 200 km, where and

The chemical stages (stages 1 and 2) deliver the vehicle from the surface to LMO; the electric propulsion stage (Stage 3) provides escape and interplanetary injection.

The ΔV allocation between stages is as follows: Stages 1 + 2 provide 5.87–6.61 km/s (surface to a 200-km circular Mars orbit, including all losses). Stage 3 provides the remaining 2.5–3.2 km/s (LMO to Mars escape plus Earth transfer injection), where the elevated ΔV relative to the impulsive equivalent (∼1.5 km/s for escape + ∼1.0 km/s for transfer shaping) reflects gravity losses inherent in low-thrust spiraling (Betts, 1998; Edelbaum, 1961).

3.3 Low-thrust spiral escape optimization

The third-stage Hall-effect thruster executes a low-thrust spiral escape from LMO. This section presents the trajectory analysis using Edelbaum’s approximation validated against numerical propagation. Equations 1–16 shows the different stages required to complete the mars sample return launch.

3.3.1 Edelbaum’s approximation

For a continuous low-thrust transfer between circular orbits, the ΔV is (Edelbaum, 1961):

For a co-planar spiral escape from to escape :

Equation 1 reduces to . At 200 km altitude (), this yields . Additional for Earth transfer shaping adds approximately 0.3–0.8 km/s, depending on departure geometry.

The methods of Whiffen and Sims (2002) and Lantoine and Russell (2012) are higher-fidelity alternatives to Edelbaum’s approximation, especially for the transition between near circular spiraling and hyperbolic escape, where the constant-thrust circular orbit assumption fails. These methods are recommended for phase A analysis.

3.3.2 Numerical propagation

The spiral trajectory was propagated numerically by solving the equations of motion in polar coordinates with continuous tangential thrust:where α is the thrust steering angle (tangential for maximum efficiency in the spiral phase).

Table 3 lists the electric propulsion stage parameters. The SPT-140 thruster has extensive flight heritage on commercial and scientific spacecraft (Hofer and Gallimore, 2006). Solar array sizing assumes triple-junction GaAs cells at 30% efficiency with 590 W/m² solar flux at Mars distance (1.524 AU), requiring 29.4 m² of deployed area. At 3.0 kg/m² for lightweight deployable arrays, the solar array mass is 88 kg, driving the Stage 3 dry mass to 103 kg. Figure 2 shows the resulting spiral escape trajectory and key parameters. The propellant mass is determined by the trajectory optimization described below.

FIGURE 2

3.4 Low-thrust robustness and degradation analysis

The baseline spiral escape analysis assumes that the thrust is applied continuously tangentially. In practice, several factors shorten effective thrust time and extend total escape duration.

3.4.1 Eclipse effects

At 200 km LMO, the eclipse fraction in each orbit is , which gives a maximum sunlit fraction of 68.6%. As altitude increases in the spiral, the eclipse fraction decreases, yielding an orbit-averaged sunlit fraction of ∼78% over a full escape trajectory.

3.4.2 Thrust duty cycle

Accounting for eclipse, thruster warm-up transients, and navigation updates, it is estimated that the effective thrust duty cycle is 85%, which would have boosted the spiral escape time from the continuous thrust value of 232 days to approximately 273 days (9.1 months).

3.4.3 Solar array degradation

Mars dust buildup and radiation damage limit solar array production. On the assumption that 2% power is lost per year due to combined effects (Zimmermann et al., 2013), this indicates that the power at the end of the spiral (9 months) is approximately 98.5% of that at the beginning-of-spiral values. This lowers thrust by exactly the same factor, increasing escape time by approximately 4 days.

3.4.4 Mars J₂ perturbation

The Mars J₂ oblateness coefficient (1.96 × 10⁻³) is responsible for the orbit plane precession of approximately 6.2°/day for the baseline of 200 km orbit (Curtis, 2020). While this does not have a meaningful effect on escape ΔV for low-inclination orbits, this restricts departure geometry and must be considered when mission planning is done.

3.4.5 Hall thruster lifetime

The SPT-140 class thruster has been tested on the ground for more than 10,000 h (Hofer and Gallimore, 2006). The spiral escape requires approximately 5,600 h of thrust time (273 days*0.85 duty cycle*24 h). This yields 56% demonstration lifetime, and there is sufficient margin for channel erosion and degradation of the cathode.

3.4.6 Sensitivity to thrust degradation

A parametric analysis was carried out for thrust from 80% to 100% of nominal (0.204 N). Table 4 shows that despite the high specific impulse, even a 20% reduction in thrust has increased the escape duration by 69 days but had only an effect of 1.2 kg on the xenon consumption. The main effect on the mission is on the mission timeline as opposed to the propellant budget. All cases are within the demonstrated thruster lifetime of 10,000 h.

The updated baseline parameters considering eclipse and duty cycle are the following: escape duration = 273 days, xenon consumption = 49.0 kg (increased from 42.0 kg to include duty cycle effects and trajectory shaping margin), and total Stage 3 propellant allocation = 52 kg (including a 6% margin).

3.5 Interplanetary return trajectory analysis

Following the low-thrust spiral escape from Mars (Sections 3.3-3.5), the spacecraft traverses the Mars sphere of influence (SOI, approx. 577,000 km radius) with a heliocentric velocity that is a function of departure geometry, as well as the excess velocity built up during the last spiral orbits. The subsequent Mars–Earth heliocentric transfer is the longest single mission phase (at approx. 12 months) and has a major impact on the total ΔV budget and mission timeline.

3.5.1 Baseline ballistic return

In the baseline design, the Hall thruster is used only for the Mars spiral escape LMO-to-SOI crossing. The heliocentric return is a ballistic Type I or Type II. For the 2031–2032 return window (corresponding to Mars departure at approximately Month 20), patched-conic analysis gives Mars departure of 2.5–3.0 km/s, and Earth arrival of approximately 4.5–5.5 km/s, corresponding to Earth atmospheric entry speed of 12.0–12.8 km/s. These values are within the envelope of heritage of the Stardust sample return capsule [12.9 km/s entry speed (Yada et al., 2021)] and are within the capabilities of the PICA-class thermal protection system (TPS) for the ERV.

3.5.2 Return trajectory ΔV partitioning

The ΔV for the complete return is partitioned as follows: low-thrust Mars spiral escape: approximately 2.80 km/s (Table 8); heliocentric transfer shaping (incorporated in spiral escape terminal targeting): approximately 0.2–0.4 km/s; midcourse corrections during the ballistic coast: approximately 0.10–0.20 km/s from the onboard chemical RCS. Earth approach targeting does not require a specific maneuver; the ERV separates from Stage 3 before Earth SOI entry and makes direct atmospheric entry. The total return-phase ΔV of approximately 3.1–3.4 km/s is consistent with the EP stage allocation provided in Table 2.

3.5.3 EP-assisted return trade study

An alternative to the ballistic baseline extends Hall thruster operation into the heliocentric phase in Table 5. Three options were evaluated parametrically:

3.5.4 Increased-fidelity analysis requirements

The patched-conic treatment of the heliocentric transfer is a first approximation. Phase A trajectory design should use direct collocation (Betts, 1998) or hybrid differential dynamic programming (Lantoine and Russell, 2012) with complete ephemeris propagation, solar radiation pressure perturbations, and finite burn modeling. The optimal low-thrust Earth–Mars transfer analyses of Mengali and Quarta (2008) are relevant analytical benchmarks, and operational experience from the Dawn mission (Rayman et al., 2006) indicates the feasibility of multi-year EP operations of the inner solar system. The trajectory optimization survey of Conway (2010) lists the applicable methods.

4 Entry, descent, and landing simulation

4.1 EDL simulation model

A three-degree-of-freedom (3-DOF) trajectory simulation was developed to assess landing accuracy and propellant requirements. The simulation models the entry vehicle as a point mass with aerodynamic forces and propulsive thrust in a rotating Mars-fixed reference frame. Bank angle modulation guidance based on Apollo-heritage predictor-corrector logic is employed during the hypersonic phase to steer the vehicle toward the target landing site, while a fuel-optimal powered descent guidance algorithm commands the propulsive terminal phase below 2 km altitude.

4.1.1 Equations of motion

In the Mars-centered, Mars-fixed spherical coordinate system (r, λ, φ), where λ is longitude and φ is latitude:where is the velocity magnitude, is the flight path angle, is the heading angle, is the bank angle, is the drag force, is the lift force, is the gravitational acceleration, is the Mars rotation rate, is the thrust magnitude, and is the thrust vector angle.

4.1.2 Aerodynamic model

The capsule aerodynamic database uses the Apollo-class vehicle characteristics scaled for Mars atmospheric conditions. Drag and lift forces are computed from standard relations , where (3.7 m base diameter) and = 0.24 (offset center of mass).

4.1.3 Atmospheric model

Mars atmospheric density follows an exponential model calibrated to Mars-GRAM (Justus and Johnson, 2001):with at the surface (Jezero elevation) and scale height km.

4.2 Monte Carlo dispersion analysis

A 10,000-case Monte Carlo simulation was performed to assess landing dispersion. Table 6 lists the dispersed parameters and their distributions. Table 6 defines the stochastic parameters used in the Monte Carlo simulation. Distributions are Gaussian, truncated at 3σ. Entry state dispersions are consistent with interplanetary navigation accuracy achievable with X-band Doppler tracking (Folkner et al., 1997). Atmospheric density uncertainty reflects seasonal and diurnal variations at Jezero Crater (Justus and Johnson, 2001). The propulsive descent corresponds to NTO/MMH engines at vacuum conditions.

In Figure 3, the scatter plot shows individual landing points in target-relative coordinates (downrange error, cross-range error), with 1σ, 2σ, and 99th-percentile dispersion ellipses overlaid. Axes are scaled in kilometers relative to the target point. Marginal histograms show that downrange and cross-range distributions are approximately Gaussian. The 99th-percentile landing ellipse measures 8.7 km × 4.2 km, sufficient for landing within the operational range of Perseverance’s cached sample tubes in Jezero Crater.

FIGURE 3

In Figure 4, peak deceleration of 12.3 g occurs at approximately 90 s; peak convective heat flux of 62 W/cm² occurs at 85 s. Propulsive descent initiates at 2 km altitude and consumes approximately 1,850 kg of NTO/MMH propellant.

FIGURE 4

Figure 5 shows the histogram of EDL propellant consumption across 10,000 Monte Carlo cases. The mean propulsive-phase propellant consumption is 755 kg; the 99th percentile is 820 kg. Including aerodynamic deceleration phase propellant and system reserves, the total EDL propellant allocation of 2,400 kg provides a substantial margin. The total EDL propellant budget includes 755 kg (mean propulsive descent) plus 1,095 kg for pre-landing maneuvers and margin.

FIGURE 5

Table 7 summarizes the key EDL performance metrics from the Monte Carlo simulation. The landing ellipse is sufficient for accessing Perseverance’s cached samples. The peak heat flux of 85 W/cm² (99th percentile) is within the capability of PICA-class thermal protection materials (Tran et al., 1997). The propellant budget of 2,400 kg is allocated in the mass budget (Section 8).

4.3 Six-degrees-of-freedom sensitivity assessment

The 3-DOF simulation shown above captures translational dynamics but not rotational dynamics, reaction control system (RCS) interactions, and attitude-dependent aerodynamic coupling. A qualitative evaluation of the 6-DOF effect on the most important EDL performance metrics is presented.

4.3.1 Attitude dynamics

For a lifting entry vehicle with L/D = 0.24, bank angle modulation requires roll control authority. This effect is to some extent captured in the bank angle execution error of ±2 deg (3 s) of Monte Carlo analysis. However, 6-DOF coupling between the pitch oscillations and drag modulation can provide a 15%–30% increase in dispersions for landing compared to 3-DOF predictions for capsule-class vehicles (Braun and Manning, 2007). Applying 30% inflation factor to the 3-DOF landing ellipse, a corrected 99th percentile landing ellipse of 11.3 km × 5.5 km is obtained, which is within the one-way, quadruped robot operational radius of 4.6 km for sample tubes within 5 km of the landing site.

4.3.2 Supersonic retropropulsion

For vehicles above a certain size (landed mass of approximately 2,000 kg), supersonic retropropulsion (SRP) is needed to obtain vehicles from supersonic to subsonic velocities (Braun and Manning, 2007; Edquist et al., 2009). SRP generates complex interactions between the bow shock and the vehicle plume that alter the vehicle’s drag coefficient and can create lateral forces. Recent computational and experimental studies of SRP aerodynamics (Berry et al., 2011) show a variation of the drag coefficient during powered flight of up to ±20% compared to unpowered values. The current simulation does not simulate this interaction. A conservative increase in propulsive descent fuel allocation of 20% is made to the mass budget to account for SRP aerodynamic uncertainty, increasing the amount of EDL propellant from 2,400 kg to 2,650 kg.

4.3.3 Plume–surface interaction

During terminal descent below approximately 100 m altitude, interaction of engine exhaust with the surface of Mars can result in the formation of craters, lofting of debris, and amplification of thrust in the ground effect. Viking, Phoenix, and InSight landing data show that these effects are manageable for engine thrust levels below 30 kN (Way et al., 2014). The terminal descent thrust of the proposed vehicle of ∼30–40 kN (for 11,800 kg at 1.5 g Mars deceleration) is at the upper end of heritage experience, which will require dedicated analysis.

4.3.4 Thermal protection system mass

Thermal protection system mass is predicted using the Sutton–Graves stagnation point heating correlation (Sutton and Graves, 1971) integrated over the entry trajectory. For the 99th-percentile heat load of 31 MJ/m² and PICA-class ablator [density = 270 kg/m³, effective heat of ablation ∼25 MJ/kg (Tran et al., 1997)], the ablator mass consumed is approximately 18.6 kg for the 15 m² wetted area. Including initial ablator thickness margin, bondline insulation, and structural backing, total TPS mass is estimated to be 350–500 kg (Table 8). This mass is allocated in the 3,000 kg capsule structure budget; however, TPS mass growth beyond 500 kg is a tracked mass risk. The Sutton–Graves correlation yields first-order estimation; computational fluid dynamics in a CO₂ atmosphere coupled with ablation response modeling [e.g., FIAT (Chen and Milos, 1999)] is required for TPS design on the flight level.

The TPS mass of 450 kg is within the 3,000 kg capsule structure allocation, which also includes the primary structure ∼1,800 kg, mechanisms ∼300 kg, and subsystem accommodation ∼450 kg. These values have been tabulated in Table 8. However, such an allocation leaves little growth margin for the capsule subsystem.

5 MA-ERV design and performance

5.1 Configuration

The MA-ERV is a three-stage vehicle: 1.3 m diameter × 5.2 m height, constrained to fit within the capsule’s central structural tube. Figure 6 illustrates the configuration.

FIGURE 6

Stage 1 (liquid bipropellant) occupies the lower 3.0 m, Stage 2 (solid motor) occupies 1.0 m, Stage 3 (electric propulsion) occupies 0.5 m, and the ERV occupies the upper 0.7 m. Total MA-ERV mass: 6,785 kg (wet).

5.2 Mass-consistent stage design

The MA-ERV mass budget is derived self-consistently using the Tsiolkovsky rocket equation applied sequentially from the top stage downward. All propellant masses, structural masses, and ΔV values are mutually consistent.

Stage 2 (Solid). Propellant formulation: Ammonium perchlorate/aluminum/PBAN [Space Shuttle SRB heritage (Sutton and Biblarz, 2017)]. I_sp = 268 s. Density = 1,750 kg/m³. The 1.3 m diameter, 0.9 m grain height envelope (subtracting the bore and nozzle cavity volumes) yields an effective grain volume of 1.032 m³ and propellant mass M_2,prop = 1,806 kg. With a nominal structural mass fraction fs = 0.10 [composite casing; range 0.08 optimistic to 0.12 conservative (Sutton and Biblarz, 2017; Humble et al., 1995)], the structural mass is M_2,struct = 181 kg. The Tsiolkovsky equation yields ΔV₂ = 3,526 m/s for an upper stage mass of 575 kg. Detailed numerical verification is provided in Supplementary Appendix A.

Stage 1 (Liquid). NTO/MMH propellant, pressure-fed cycle, I_sp = 311 s [vacuum, heritage engines (Huzel and Huang, 1992; ArianeGroup, 2018)]. Using the effective propellant volume in the 1.3-m diameter, 3.0-m height first-stage envelope, the effective propellant volume is approximately 3.25 m³: yielding M_1,prop = 3,803 kg at NTO/MMH propellant bulk density of 1,170 kg/m³ (mixture ratio 1.65:1). The value of M_1,struct for structural mass fraction f_s = 0.08 is 304 kg. Using the Tsiolkovsky equation, we find ΔV = 2,647 m/s. Combined chemical ΔV: ΔV₁ + ΔV₂ = 6.17 km/s. Detailed derivations are provided in Supplementary Appendix A.

Table 9 presents the self-consistent mass and performance breakdown for each MA-ERV stage. All values satisfy the Tsiolkovsky equation simultaneously. The combined chemical ΔV of 6.17 km/s exceeds the required 5.87 km/s (lower bound from Table 2) by 300 m/s, providing a limited margin for trajectory dispersions. This reduced margin compared to the original grain-volume estimate underscores the sensitivity of architecture closure to Stage 2 propellant loading. Total MA-ERV wet mass: 4,107 kg + 1,951 kg + 145 kg + 350 kg = 6,553 kg (rounded to 6,600 kg for budgeting).

5.3 First-stage engine trade study

Six engine options were evaluated. Table 10 summarizes engine candidates evaluated for the MA-ERV first stage. Options A–D leverage heritage or minimally modified engines; Options E–F require new development. The SuperDraco vacuum of 270 s is estimated from its sea-level value of 235 s (Seedhouse, 2015) using nozzle expansion corrections; published vacuum data are unavailable. The OMS-E and AESTUS nozzles are truncated to fit the 3.0 m height constraint, with performance decremented per Sutton and Biblarz (2017).

A weighted multi-attribute utility analysis (Wertz and Larson, 1999) was conducted. Table 11 shows the multi-attribute trade study results. Criteria weights were determined by pairwise comparison following the methodology of Wertz and Larson (1999). Option A (all-solid, first stage) scores highest overall due to strong ΔV margin and cost advantages. Option B (SuperDraco) is recommended for configurations prioritizing schedule and EDL dual-use.

5.4 Parametric sensitivity analysis

Figure 7 shows the sensitivity of combined chemical to the three most uncertain parameters.

FIGURE 7

5.5 Structural load analysis and thrust-to-weight assessment

5.5.1 Thrust-to-weight ratio

The MA-ERV must have a sufficient thrust-to-weight ratio (T/W) at liftoff from Mars to minimize the losses due to gravity. A baseline MA-ERV with an initial mass of 6,553 kg requires approximately 49 kN of thrust to achieve T/W = 2.0 on Mars (g_Mars = 3.72 m/s²). The corresponding propellant mass flow rate of 16.1 kg/s (at I_sp = 311 s) results in a Stage 1 burn time of 236 s. T/W changes from 2.01 at liftoff to 4.79 at Stage 1 burnout with constant-thrust propellant burn. Following a 5-s staging coast, Stage 2 fires with an initial value of T/W = 5.50 to a final value of approximately 13.3 at burnout (Figure 8).

FIGURE 8

At Stage 1 burnout, T/W increases to , which is acceptable. Stage 2 ignition at 2,446 kg initial mass with SRM thrust of approximately 50 kN yields T/W = , which is appropriate for an upper stage in near-vacuum conditions.

5.5.2 Structural mass fraction justification

The assumed 0.08 (nominal) structural mass fraction requires questioning.

For Stage 1 (pressure-fed liquid): Thin-wall pressure vessel analysis for titanium (Ti-6Al-4V) tanks at 20 bar feed pressure gives a tank mass of ∼16 kg, which is only 0.4% propellant mass. However, total structural mass consists of a feed system, valves, engine, thrust structure, interstage, wiring, and insulation. Historical data for pressure-fed systems of this class (Sutton and Biblarz, 2017; Wertz et al., 2011) point to achievable structural fractions of 0.09–0.15 [the lower bound requiring composite overwrap pressure vessels (COPVs)].

For Stage 2 (solid motor), the structural fraction includes casing, insulation, nozzle, and igniter. Composite-cased SRMs (e.g., Orbus-class) achieve 0.08–0.09 (Humble et al., 1995), while steel-cased motors range from 0.11 to 0.15. Internal insulation thickness for a burn time of 53 s at the proposed grain regression rates requires approximately 3–5 mm of EPDM liner, adding approximately 15 kg.

5.5.3 Structural fraction assessment

Based on the above analysis, the nominal structural mass fraction is revised as 0.08 to 0.10 with a range of 0.08 (optimistic, COPV) to 0.13 (conservative, steel casing with thick insulation). Table 12 provides the updated performance of MA-ERV with f_s = 0.10.

5.5.4 Ascent trajectory losses

The gravity loss estimate of 0.80–1.20 km/s is refined using the following relationship (Wertz et al., 2011):For a Stage 1 burn time of 236 s, the average flight path angle is 60°, and Mars gravity is as follows: . This is consistent with the lower bound of the parametric estimate.

6 Quadruped robot system

6.1 Platform specifications

6.2 Mars energy model

A physics-based energy model was developed for quadruped locomotion on Mars. The model accounts for locomotion power, thermal management, and environmental factors.

6.2.1 Locomotion power

The power for quadruped walking on flat terrain is modeled as follows (Minetti and Alexander, 1997):where is the cost of transport (dimensionless). For Spot-class quadrupeds, on flat terrain (Seok et al., 2015), increasing with slope:where for legged robots (Hutter et al., 2012).

On Mars (², ):

6.2.2 Martian regolith interaction

Terrestrial COT values must be corrected for the properties of the Martian regolith, which varies from compacted basalt to loose aeolian sand (Moore and Jakosky, 1989). Using the Bekker–Wong terramechanics framework (Wong, 2008), we can compute that the ground pressure of the 35 kg robot (50 cm² foot contact area) is approximately 6.5 kPa. On compacted regolith [bearing capacity >20 kPa (Moore and Jakosky, 1989)], there is little sinkage. On loose sand (bearing capacity ∼5–10 kPa) with projected sinkage of 1–3 cm, C_OT is increased by 15%–30% based on terrestrial legged robot data (Li et al., 2013). Using a terrain-averaged regolith penalty factor of 1.20, the flat terrain Mars C_OT is increased from 2.0 to 2.4, with corresponding proportional range reduction. The corrected Mars COT with slope effects is .

6.2.3 Thermal power

Mars surface temperatures of −60 °C to −120 °C require active heating of the robot. For an insulated robot body (0.5 m² effective surface area, 3 cm aerogel insulation at K = 0.015 W/mK), the thermal resistance is 4.0 K/W, providing a heating requirement of approximately 20 W at −60 °C ambient temperature. This thermal load is factored into the total power model together with the locomotion, avionics, and communications power.

Figure 9 shows the energy analysis of the quadruped robot Mars energy.

FIGURE 9

Figure 10 demonstrates the sample retrieval sortie simulation.

FIGURE 10

Upper panel: instantaneous power consumption and cumulative distance over a complete sortie cycle (deploy → traverse out → search/pickup → traverse back → delivery). The peak power of ∼420 W occurs during traverse phases at 0.8 m/s over 5° average slopes.

Lower panel: combined battery and fuel cell state of charge versus time. The system reaches 10% reserve threshold at approximately 6.8 h, consistent with the analytical predictions in Table 13.

6.3 Fuel cell range extension

A methane–oxygen fuel cell system is integrated to extend operational range beyond battery-only capability. The fuel cell operates via the reverse of the Sabatier reaction:

Table 13 derives the total energy available from the onboard methane fuel cell system. The resulting operational time extension enables sample retrieval sorties with a radius of approximately 4.6 km from the capsule (assuming 0.8 m/s average traverse speed and 50% of energy allocated to the return trip).

6.4 Sample delivery system

Two complementary methods transport samples to the ERV:

6.4.1 Method 1: conveyor system (for sample tubes)

Sample tubes received at a port near the capsule base are transported by a two-stage conveyor belt (horizontal, then vertical) to the top of the capsule, where dual robotic arms transfer them to the ERV sample compartment.

6.4.2 samples Method 2: ladder climbing (for irregular)

A foldable four-stage ladder deploys from the former parachute compartment, enabling quadruped robots to climb to the capsule top and directly deliver samples to the robotic arms. The ladder climbing relies on groove features machined into the robot’s leg contact surfaces that interlock with ladder rungs. In Mars’s 0.38 g environment, the required joint torques are approximately 38% of Earth values.

6.4.3 Ladder climbing feasibility

In the Mars 0.38 g environment, the joint torques required are approximately 38% of those on Earth. Static stability for a quadruped on a 70-degree ladder (robot mass = 35 kg, payload = 2 kg) yields a positive stability margin of 6.5 Nm, which is a factor of safety against quasi-static tipover. The critical angle at which the static stability disappears is 75° (Figure 11a). The required hip joint torque at 70° is approximately 14 Nm per leg, which is 35% of the estimated 40 Nm actuator limit. The climb energy cost of 0.83 Wh is insignificant compared with the 2,915 Wh sortie energy budget. Derivations are given in detail in Supplementary Appendix A.

FIGURE 11

Figure 11b shows the longitudinal static stability margin versus ladder angle for Mars (0.38 g) and Earth (1 g). The critical angle where static stability vanishes is 75°; above this angle, active stabilization is required. The design angle of 75° lies at the stability boundary. Figure 11b shows the required hip joint torque per leg during the tripod climbing gait. On Mars, the torque at 75° is approximately 15 N·m, well within the estimated Spot actuator limit of 40 N·m (37% utilization). Figure 11c shows the mechanical power required for climbing at 0.1 m/s. Mars climbing power is 38% of Earth’s values. Figure 11d shows the time and energy to climb the 4.5 m capsule height at various ladder angles. At 75° and 0.1 m/s climb speed, the path length is 4.5/sin (75°) = 4.66 m, yielding a climb time of 46.6 s. At approximately 60 W Mars climbing power (from Figure 11c), the energy consumed is 0.78 Wh, which is negligible compared to the 2,915 Wh sortie energy budget.

6.5 Surface communications architecture

Autonomous sample retrieval up to a distance of 4.6 km from the lander requires a robust communications link on the surface. Three modes of communication are envisaged:

6.5.1 Direct robot-to-lander UHF link

Over relatively flat terrain in the Jezero Crater, at a distance of 4.6 km, line-of-sight UHF communications at 2 W transmit power offers data rates of ∼8–32 kbps, which are adequate for telemetry and low-bandwidth command. Signal attenuation caused by dust storms may decrease the range to approximately 3 km.

6.5.2 Orbiter relay

Mars Reconnaissance Orbiter (MRO) or a future relay orbiter: 2–4 contact windows are available each sol for high bandwidth return of data and upload of commands. Latency of 4–24 min (one-way, light time) prohibits real-time control and requires autonomous navigation on board.

6.5.3 Inter-robot mesh network

When deployed concurrently, the three robots can act as relay nodes, expanding the effective communication range. This mode requires software-defined networking functions that are not yet standard on the Spot platform.

The predominant mode of communication during nominal operations is direct-to-lander UHF. On each robot is a 2 W UHF transceiver (0.3 kg mass, 5 W electrical power) with a hemispherical antenna. The power of the communication 5 W is contained in the term of Equation 13.

Autonomous navigation is based on stereo vision, IMU dead-reckoning, visual odometry, and Sun-sensor absolute heading correction. The design of the autonomous navigation system is an adaptation of the method used by the Perseverance rover’s AutoNav system (Maurette, 2003), which is scaled for legged locomotion. This capability is TRL 3 for quadruped platforms and requires dedicated development and Mars-analog field testing.

7 ISRU and energy systems

7.1 ISRU system design

The ISRU system produces methane and oxygen via the Sabatier reaction (NASA, 2016; Battin, 1999):followed by water electrolysis to recycle hydrogen: is acquired from the Martian atmosphere [95.3% by volume (Justus and Johnson, 2001)]. Hydrogen is brought from Earth as water or liquid hydrogen (initial supply), with subsequent recycling from fuel cell water products. Table 14 presents the ISRU system power budget.

Table 14 compares ISRU subsystem power consumption across three design points. The “Laboratory” column reflects the as-built prototype data from Zubrin et al. (2013). The “Optimized” column applies mass and power reductions projected for a flight design. The “With RTG Heat” column uses waste heat from the general purpose heat source–radioisotope thermoelectric generators (GPHS-RTGs) to drive thermal and mechanical loads, reducing electrical demand from 678 W to 274 W per unit. The electrical demand of 274 W per ISRU unit is provided by the RTG electrical output and wind turbine system.

7.2 RTG power and thermal management

The thermal power serves multiple functions: (1) spacecraft thermal management during transit and on Mars surface (−60 °C to −120 °C ambient), (2) driving ISRU thermal and mechanical loads as analyzed above, and (3) heating the quadruped robots during recharging.

7.2.1 Thermodynamic justification for waste heat utilization

The RTG hot side temperature of approximately 1,273 K and cold side temperature of approximately 300 K result in a Carnot efficiency limit of 76.4%. A realistic Stirling or thermoacoustic converter that is 35%–40% Carnot efficient converts 27%–31% thermal to mechanical (Singhal and Kendall, 2003). The required 2276 W of mechanical work from 8800 W of waste heat is 25.9% conversion efficiency, well within the range.

7.2.2 Heat engine mass and integration

Conversion of RTG waste heat to mechanical work for ISRU thermal/mechanical loads requires a Stirling or thermoacoustic converter. Based on the development of the Stirling convertor by the US National Aeronautics and Space Administration’s Glenn Research Center (NASA Glenn) (Oriti, 2015), a 2.3 kW mechanical output power level Stirling engine is estimated to be 20–30 kg in mass (specific power ∼75–115 W/kg at this scale). Adding thermal coupling hardware (heat exchangers, plumbing), the total heat-to-work conversion system mass is estimated at 35–50 kg. This mass is added to the system mass budget.

The heat engine integration calls for thermal coupling between the RTG hot shoe (∼1,273 K) and the Stirling acceptor. This interface must be able to accommodate thermal expansion differentials and maintain seal integrity over the ∼11 months surface mission duration. Heritage from the Advanced Stirling Radioisotope Generator (ASRG) program (NASA Glenn Research Center, 2013) provides relevant design data, although ASRG was canceled before qualification for flight (TRL 5–6 at program termination).

7.3 Wind turbine system

The power output of a horizontal-axis wind turbine is , where ρ is the air density, is the wind speed, is the rotor radius, and is the power coefficient. The Betz limit yields (Betz, 1920); a realistic value for small turbines is (Lichter and Viterna, 1999). Table 15 presents electrical output at Jezero Crater conditions for turbines with a rotor radius of approximately 1.9 m (3.8 m diameter). The 15 kg per-turbine mass allocation is aggressive and requires structural validation for a 3.8 m deployable rotor operating in dust storm conditions.

At an annual mean wind speed of 7–9 m/s (Holstein-Rathlou et al., 2010), wind power supplements but cannot fully substitute for RTG electrical output. The full power budget (1,526 W) closes at wind speeds above approximately 12 m/s. At the mean wind speed, ISRU operations must be time-shared with other power demands. During dust storms (wind speeds 20–30 m/s), substantial excess power is available, although turbine survival in these conditions requires structural reinforcement. Figure 12 shows the Mars wind turbine and ISRU system analysis.

FIGURE 12

7.3.1 Dust mitigation and survivability

Mars wind turbine operation is challenged by three dust aspects: dust abrasion of blade surfaces, dust buildup on mechanical components, and electrostatic dust adhesion. Blade erosion can be overcome by ceramic coatings or sacrificial leading-edge strips. Turbine hub and generator bearing seals rated for Mars dust [particle size 1–40 mm (Justus and Johnson, 2001)] are required.

The mass allocation for wind turbines is updated from 15 kg to 22 kg for each wind turbine to accommodate dust-hardened bearings (2 kg), blade protective coatings (1 kg), sealed generator housing (2 kg), and deployment mechanism reinforcement (2 kg). Total wind turbine system mass: 4 × 22 kg = 88 kg.

7.3.2 Power-time scheduling strategy

For the RTG, at mean wind speeds of 7–9 m/s, combined RTG + wind turbine power (approximately 660–726 W) is not enough to run all four ISRU units simultaneously (needs 4 × 274 W = 1,096 W electrical). The operational strategy uses time division scheduling:

Priority 1 (continuous): Spacecraft thermal management, robot battery charging during overnight periods: ∼200 W from RTG

Priority 2 (opportunistic): ISRU production during sufficient wind operates 1–2 units at 7–9 m/s, all four units above 15 m/s

Priority 3 (burst): Full ISRU production during high-wind events (>15 m/s): leverages cubic wind power scaling

At Jezero Crater, seasonal wind speed data from Mars-GRAM (Justus and Johnson, 2001) show that wind speeds more than 12 m/s are expected 20%–30% of the time in the surface operations window, and more than 15 m/s of the time 10%–15% of the time. Methane production adequate to fuel all three robots for only one full sortie each (total ∼1.25 kg CH₄) takes approximately 60–120 h of ISRU operation, depending on wind conditions, and is easily possible within the framework of 11 months of surface operations.

In the fallback scenario where ISRU production is insufficient (e.g., extended low-wind period), the robots operate in battery-only mode with a reduced range of 1.2 km per sortie. This limits sample retrieval to tubes within approximately 1 km of the landing site, which is estimated at 5–10 tubes based on Perseverance’s caching pattern in Jezero Crater (Muirhead et al., 2020a).

8 Mass and cost estimation

8.1 Mass budget with subsystem mass growth allowances

Table 16 shows the updated mass budget using AIAA S-120A compliant mass growth allowances (MGAs) (AIAA, 2015) assigned per subsystem as a function of current TRL and design maturity. Unlike the previous method of using a single system-level margin, this methodology allocates greater growth allowances to less mature subsystems, as is consistent with historical mass growth patterns for space missions.

The revised mass budget indicates that there is a deficit of approximately 1,010 kg (6.0%) compared to Falcon Heavy expendable capacity. This finding is significant: it shows that the architecture does not close on the baseline Falcon Heavy launcher when the proper subsystem-level mass growth allowances are made. The previous analysis with a single 20% system margin resulted in apparent closure due to underweighting of the growth risk from the lowest-TRL subsystems.

8.1.1 Probabilistic mass analysis

A 5,000-case Monte Carlo analysis was conducted using lognormal distributions for each subsystem, chosen over Gaussian distributions because: (1) mass cannot be negative, (2) aerospace subsystem mass growth is empirically positively skewed (Thompson et al., 2010), and (3) multiplicative growth factors produce lognormal outcomes. Each subsystem mass and is calibrated so that the maximum expected value falls at the 70th percentile, consistent with the AIAA S-120A MEV definition (AIAA, 2015). Figure 13 shows the resulting distribution.

FIGURE 13

The likelihood of closing within Falcon Heavy capacity is roughly 38% under these assumptions, which is considerably lower than the 87% calculated using symmetric Gaussian distributions. This reinforces the fact that the architecture requires either:

8.1.1.1 Mass reduction

Elimination of ISRU, wind turbines, and heat engine systems (saving ∼390 kg CBE, ∼530 kg MEV) achieves a simplified architecture of battery-only robot operation and a 1.2-km retrieval radius.

8.1.1.2 Higher-capacity launcher

Falcon Heavy with performance upgrades (∼18,000 kg class), Vulcan Centaur Heavy (∼18,500 kg Mars transfer), New Glenn ∼17,500 kg estimated, or space launch system (SLS) Block 1 (∼27,000 kg, substantial margin).

8.1.1.3 Subsystem mass reduction

Specific technology development aimed at a capsule structure mass of 2,500 kg (composite primary structure) and f_s = 0.08 for MA-ERV stages through composite casing could result in a total CBE reduction of ∼800 kg.

To quantify the probabilistic implications of these differences, a 5,000-case Monte Carlo mass simulation was conducted using lognormal distributions for each subsystem, replacing the symmetric Gaussian model employed in the original analysis. Table 17 shows the mass closure probability by launcher.

8.1.2 Probabilistic modeling methodology

Each subsystem mass is modeled as a lognormal random variable:

, where and is calibrated such that the maximum expected value [MEV = CBE × (1 + MGA)] falls at the 70th percentile of the distribution, consistent with the standard MEV definition in AIAA S-120A (AIAA, 2015). The lognormal distribution is chosen over the Gaussian distribution for three physically motivated reasons: (1) mass cannot be negative, providing a natural lower bound; (2) aerospace subsystem mass growth is empirically observed to be positively skewed, with growth exceedances substantially more likely than under-runs (Thompson et al., 2010); and (3) the multiplicative nature of mass growth factors (structural fraction uncertainties, design changes, and qualification additions) produces log-normally distributed outcomes by the central limit theorem applied to products. The total system mass is computed as the sum of all subsystem samples plus a deterministic 10% system-level margin applied to the summed MEV, representing irreducible integration and harness mass that scales with overall system size.

8.2 System mass sensitivity analysis

Figure 13a shows a horizontal bar chart decomposing each subsystem into its current best estimate (CBE, blue) and mass growth allowance (MGA, orange) components. The MGA magnitudes are assigned per AIAA S-120A guidelines based on assessed technology readiness level: 30% for TRL 3 subsystems (capsule structure, quadruped robots, ISRU, and wind turbines), 25% for TRL 4 (MA-ERV Stage 1, ERV), 20% for TRL 5 (MA-ERV Stages 2–3, service module), 15% for TRL 6 (communications), and 5% for TRL 7–8 (RTGs, EDL propellant). The MA-ERV Stage 1 and capsule structure dominate both absolute mass and growth allowance, together contributing more than 60% of total MGA.

Figure 13b illustrates the histogram of total launch mass (including 10% system margin on MEV) across 5,000 lognormal Monte Carlo samples. The distribution exhibits the characteristic positive skew of lognormal sums, with a long right tail extending beyond 19,000 kg. The mean total mass is approximately 17,810 kg (black vertical line). The Falcon Heavy expendable capacity of 16,800 kg (dashed red line) intersects the distribution at the 38th percentile, indicating that only 38% of Monte Carlo cases close within this launcher’s capability. The Vulcan Centaur Heavy capacity of 18,500 kg (dashed green line) intersects at the 78th percentile, providing substantially higher closure confidence. The contrast between the 38% lognormal closure probability and the 87% Gaussian closure probability reported in the original analysis quantitatively demonstrates the reviewer’s concern that symmetric uncertainty models substantially overestimate closure confidence for architectures with multiple low-TRL subsystems.

Figure 13c presents the mass allocation pie chart based on CBE values, grouped by functional category. Propulsion elements (MA-ERV Stages 1–3, ERV, and EDL propellant) collectively account for 57% of the total CBE mass, confirming that the architecture is fundamentally propulsion-dominated. The capsule structure (18%) and power/surface systems (3%) represent secondary but non-negligible contributors. The 10% system margin (applied to MEV, not shown in CBE breakdown) adds approximately 1,620 kg.

Figure 13d shows the closure probability as a continuous function of launcher capacity, computed by evaluating the cumulative distribution function of the Monte Carlo total mass distribution at each capacity value. Discrete markers indicate the assessed closure probabilities for four launcher options: Falcon Heavy expendable (16,800 kg, 38%), New Glenn estimated (17,500 kg, 55%), Vulcan Centaur Heavy (18,500 kg, 78%), and SLS Block 1 (27,000 kg, >99%, off-scale). The horizontal dashed line at 80% represents a notional programmatic threshold for proceeding past Mission Concept Review; only the Vulcan Centaur Heavy approaches this threshold among near-term commercially available launchers. The steep slope of the curve between 16,000 kg and 19,000 kg indicates that relatively modest mass reductions (∼800–1,000 kg through composite structures and optimized propulsion) could substantially improve closure probability on the Falcon Heavy class of launchers.

8.3 Cost estimation

Table 18 shows updated cost estimates based on more conservative assumptions for low-TRL subsystems, historical cost growth factors, and a strengthened comparison with similar missions. The methodology employs the use of the following: NASA NAFCOM parametric weight-based cost estimating relationships (CERs) (NASA, 2015) for spacecraft hardware with adjustments for commercial procurement where applicable. A total of $3.0–6.0 billion reflects more conservative treatment of technology development costs, especially for the MA-ERV ($200–500 million vs. the previous $50–150 million) and capsule modification ($500–900 million vs. $300–600 million). The cost increase for the MA-ERV is warranted by comparison with the development of small launch vehicles: the development of Rocket Lab’s Electron rocket cost approximately $200 million for a much simpler single-stage vehicle, while multi-stage solid/liquid/electric propulsion integration at the required reliability for MSR is unprecedented.

8.3.1 Historical comparison

The Mars Science Laboratory (Curiosity) mission cost was $2.5 billion for a 900 kg rover (Way et al., 2014). Mars 2020 (Perseverance) cost: approximately $2.7 billion. The proposed architecture, with much greater complexity (EDL for ∼12,000 kg, MA-ERV ascent, EP escape, robotic sample retrieval, and Earth return), would be expected to cost much more on a “per subsystem” basis. The estimated range of $3.0–6.0 billion reflects this complexity while allowing for the reduction in costs from single-launch consolidation and the use of commercial hardware. However, cost estimates at the conceptual design stage contain factors of uncertainty of 2–3 times (NASA, 2015), and the upper limit should not be regarded as a firm limit.

9 Discussion

9.1 Feasibility assessment

This architecture operates at the conceptual design level (NASA Pre-Phase A equivalent). The analyses presented establish physical plausibility but do not constitute engineering-level design verification. Table 19 identifies critical technology gaps. The 2029 launch window is unrealistic given TRL gaps of 3–4 for multiple subsystems; a 2035–2037 window is more appropriate, allowing 10–12 years for technology maturation, system integration, and qualification testing. The additional schedule relative to the original 2033 estimate reflects the simultaneous maturation requirement across propulsion, EDL, robotics, ISRU, and containment subsystems, which is historically unprecedented within a single program at this budget scale.

9.1.1 Technology maturation roadmap

A phased development approach is recommended:

Phase I (Years 1–4): Parallel technology maturation for highest-risk elements: capsule structural modification (finite-element analysis (FEA), structural test article), Mars propulsive EDL (SRP testing, 6-DOF simulation), and MA-ERV Stage 2 solid motor (grain design, case qualification).

Phase II (Years 3–6): Subsystem integration: MA-ERV three-stage integration testing, quadruped robot Mars-analog field trials, and ISRU pilot production system

Phase III (Years 5–8): System integration: full-scale EDL testing (high-altitude drop tests), MA-ERV hot-fire test series, and end-to-end sample retrieval demonstration

Phase IV (Years 7–10): Flight system assembly, test, and launch operations

9.2 Comparison with competing architectures

Table 20 provides a systematic comparison across key metrics. The proposed architecture offers the lowest cost and the highest sample capacity through its single-launch consolidation approach. However, it carries substantially higher technical risk than the NASA baseline due to the very tight mass margin (1.8%–3.4% versus typically >15% for flagship missions) and the requirement for several TRL 3–4 technologies to reach TRL 6–7. The mass margin is below the threshold generally accepted for proceeding past Mission Concept Review, indicating that either subsystem mass reductions or a higher-capacity launch vehicle is needed.

9.3 Risk assessment

Table 21 identifies the top mission risks. Mass budget exceedance (score 20) is the highest-rated risk, reflecting the very tight 1.8%–3.4% margin. This risk can be partially mitigated by (a) removing ISRU and wind turbine systems (simplified version, saving ∼320 kg), (b) reducing EDL propellant through trajectory optimization, or (c) using a higher-capacity launch vehicle if available.

9.3.1 Single-point-of-failure risk

The single-launch architecture inherently concentrates mission risk. Unlike the NASA/ESA baseline that distributes functions across three spacecraft (any one of which can partially succeed independently), a single EDL or ascent failure in the proposed architecture results in complete mission loss. This risk concentration is the fundamental cost of single-launch consolidation and must be weighed against the cost and schedule benefits. Partial success scenarios are limited to (a) successful landing but MA-ERV failure, which would yield surface science data from the quadruped robots but no sample return, and (b) successful ascent but EP failure, which is potentially recoverable if chemical stages achieve sufficient velocity for a direct (non-optimal) Earth-return trajectory. A quantitative mission reliability assessment, accounting for subsystem reliability allocation and single-string versus redundant architectures, is required during Phase A study.

9.4 Planetary protection compliance

Mars sample return missions fall under COSPAR Category V Restricted Earth Return (COSPAR, 2021), which imposes very strict requirements for avoiding uncontrolled release of Martian material into the Earth’s biosphere. The most important requirements and their design consequences are described in the following.

9.4.1 Triple containment

9.4.2 Break-the-chain

No hardware contacting the Martian environment unsterilized can be brought back to Earth without being contained. In the proposed architecture, the quadruped robots deal with cached sample tubes that have been sealed by the Perseverance. However, the outside surfaces of the tubes have been exposed to the Martian environment. The sample loading mechanism must either: (a) sterilize the outside of loading tubes prior to loading them into the OS container, or (b) treat the inside of the OS container as “Mars-exposed” and rely on the primary, secondary, and tertiary containment barriers.

9.4.3 ERV mass allocation

The ERV mass allocation of 420 kg tabulated in Table 22 includes the following details.

The total ERV of 420 kg is an aggressive allocation compared to the ESA Earth Return Orbiter capture and containment system [estimated >600 kg (Muirhead et al., 2020a)]. The mass savings are possible because the proposed ERV is a direct-entry capsule (no insertion into orbit at Earth), which is similar in concept to Stardust’s sample return capsule (46 kg) but scaled up for the increased sample payload and more stringent contamination requirements.

9.4.4 Abort and breach scenarios

Two critical scenarios require analysis:

9.4.4.1 Containment breach during Mars ascent

In the event of a failure of the OS container seal during MA-ERV launch vibration, samples will stay inside the ERV shell (tertiary containment). Leak detection sensors on the OS container can check the integrity of the seal in LMO before committing to return to Earth.

9.4.4.2 ERV breakup during Earth entry

For entry into Earth to occur during return from Mars, the ERV must survive entry at approximately 12 km/s (direct return from Mars). TPS failure could contaminate Earth’s atmosphere with Martian material. Mitigation factors include robust TPS design (factor of safety to ablation thickness >2.0), trajectory biasing to ensure steep entry (avoiding skip-out to uncontrolled re-entry), and targeting strategy ensuring impact within designated recovery zone [e.g., the Utah Test and Training Range (UTTR)].

Planetary protection compliance is listed as one of the key design drivers that must be the focus of dedicated analysis in Phase A. The mass, cost, and schedule impacts could be substantial and are not fully represented in the existing conceptual estimates.

9.5 Communications architecture

Surface operations that require autonomous robot coordination over a range of 4.6 km use a layered communications architecture as described in Section 6.5. The most critical issues that remain open are described in the following subsections.

9.5.1 Latency

There is a 4–24 min one-way light time between Earth and Mars, which would prevent real-time teleoperation. All sample retrieval operations are to be autonomous with ground-in-the-loop supervisory control limited to strategic decisions (sortie planning, path selection, and abort commands).

9.5.2 Data return capacity

Navigation camera imagery, health telemetry, science data ∼100 Mbits/sortie for 10 W of power; direct-to-Earth X-band, approximately 0.5–2 kbps 14–55 h per sortie for complete data return. Orbiter relay by way of MRO [0.5–2 Mbps during contact windows (∼10 min/pass, 2–4 passes/sol)] for same-sol data return.

9.5.3 Autonomous navigation reliability

Currently, the state of the art in quadruped autonomous navigation in an unstructured outdoor environment achieves >95% waypoint success rate on Earth (Boston Dynamics, 2024). Mars-specific difficulties and challenges (lighting conditions, obscuration due to dust, and novelty of terrain) need thorough simulation and analog testing.

9.6 Limitations

This study has the following significant limitations, several of which fundamentally bound the credibility of feasibility claims:

9.6.1 Pre-Phase A design maturity

This work is at the conceptual design level using parametric models and simplified analyses. Engineering-level design verification requires many more high-fidelity tools and hardware testing.

9.6.2 EDL simulation fidelity

The 3-DOF simulation is based on simplified aerodynamic and propulsion models. The 6-DOF sensitivity assessment (Section 4.3) suggests the possibility of inflating the landing ellipse by 30%. A complete 6-DOF simulation with a detailed aerodynamic database validated by computational fluid dynamics (CFD) in a CO₂ atmosphere, reaction control system modeling, and terrain interaction is required.

9.6.3 No structural analysis

Finite-element analysis of the modified capsule structure, MA-ERV, the central structural tube under ascent loads, and the mechanical interfaces is essential to validate the assumed structural mass fractions. The central tube must withstand >50 kN thrust during MA-ERV Stage 1 ignition.

9.6.4 No thermal analysis

For Mars EDL aerothermal analysis in a CO₂ atmosphere, specific computational fluid dynamics with coupled ablation modeling [e.g., FIAT (Chen and Milos, 1999)] is needed. Surface thermal management for 11 months of Mars operations: Transient thermal modeling is required.

9.6.5 Trajectory optimization is preliminary

The low-thrust spiral escape uses tangential thrust steering, and the interplanetary return is treated at the patched-conic level. A complete optimal control solution based on direct collocation (Betts, 1998), on the other hand, by hybrid differential dynamic programming (Lantoine and Russell, 2012), or by the smoothing methods of Taheri and Junkins (2018), would increase propellant efficiency by an estimated 3%–8% for the spiral phase, and could allow more stringent return trajectory design including continuous thrust arcs (Mengali and Quarta, 2008; Conway, 2010).

9.6.6 Robot locomotion model is analytical

No multibody dynamics simulation, hardware-in-the-loop testing, or reduced-gravity prototype testing has yet been carried out. The regolith interaction model uses terrestrial analogs.

9.6.7 Planetary protection analysis is preliminary

The ERV containment architecture is parametrically sized. Detailed contamination control, sterilization verification, and Earth entry survival analysis are required for COSPAR compliance.

9.6.8 Communications architecture is conceptual

No link budget analysis, antenna pattern modeling, or autonomous navigation performance prediction has been conducted.

9.6.9 Mass budget does not close on the baseline launcher

With a good subsystem MGA, the architecture overshoots Falcon Heavy by approximately 6%, requiring mass reduction or a larger launcher.

9.6.10 Cost estimate carries a factor-of-2 uncertainty

At the conceptual design stage, cost estimates are inherently unreliable and can increase considerably during detailed design.

10 Conclusion

This article has presented and quantitatively evaluated a conceptual single-launch Mars sample return architecture that combines a capsule-class entry vehicle, a three-stage Mars Ascent–Earth Return Vehicle, and a redundant quadruped robotic sample retrieval system into one consolidated spacecraft. Five major findings emerge from the analysis.

First, the three-stage MA-ERV has a combined chemical ΔV of 5.63–6.17 km/s, depending on structural mass fraction assumptions (0.10–0.08), bracketing the minimum required 5.87 km/s and thus showing that architecture closure is sensitive to structural technology. The use of the Hall-effect electric propulsion stage adds another 2.80 km/s for the Mars escape and for the Earth transfer, confirmed by numerical propagation of the spiral trajectories with an agreement within 5% with Edelbaum’s analytical prediction, with robustness analysis showing that even 20% degradation of thrust results in an increase of escape time of 69 days, without jeopardizing propellant budget closure. In addition, a newly included interplanetary return trajectory analysis verifies that the baseline ballistic coast after Mars spiral escape produces Earth entry speeds of 12.0–12.8 km/s within heritage TPS capabilities. A parametric trade study of further EP thrusting in the heliocentric phase shows that 2–3 months of transit time could be recovered at a cost of 8–15 kg of additional xenon, representing a viable trade option if relief of mass budget could be achieved.

Second, the 10,000-case Monte Carlo EDL simulation shows that a capsule-class vehicle with propulsive terminal descent provides 8.7 km × 4.2 km (3-DOF) 99th percentile landing accuracy (11.3 km × 5.5 km when 6-DOF effects are conservatively applied). Parametric TPS sizing provides 450 kg within the capsule structure allotment, with maximum heat flux inside PICA-class thermal protection conditions.

Third, the quadruped robot energy model, which accounts for Martian regolith interaction penalties, results in the triple-redundant robotic retrieval of cached sample tubes with ISRU-supplied methane–oxygen fuel cells, increasing the battery-only operational radius from 1.2 km to 4.6 km per sortie and a surface communications architecture.

Fourth, the full mass budget using AIAA S-120A compliant subsystem mass growth allowances is approximately 17,800 kg, which is approximately 6% more than the Falcon Heavy expendable capacity of 16,800 kg. This finding, in contrast to the apparent closure demonstrated by simpler margin methods, is illustrative of the fact that the architecture requires either focused subsystem mass reduction (approximately 800 kg by composite structures and optimized propulsion) or migration to a higher-capacity launch vehicle. Probabilistic analysis gives 38% closing probability on Falcon Heavy (up to 78% on Vulcan Centaur Heavy).

Fifth, an initial planetary protection compliance study identifies the ERV containment system as a key design driver at 420 kg allocation with large mass, cost, and schedule ramifications requiring dedicated Phase A analysis. The revised cost estimate of $3.0–6.0 billion reflects more conservative treatment of technology development costs and, at the same time, represents a potential reduction compared to the multi-spacecraft, as a baseline, the MSR design by the American and European space agencies (NASA/ESA).

The major weakness of the architecture is the negative mass margin under appropriate growth allowances, which is well below the ≥15% threshold requirement at Mission Concept Review for flagship missions. Seven technology elements must progress from TRL 3–4 to TRL 6–7 over an estimated 10–12 years development timeline, with capsule structural modification, Mars propulsive EDL, and planetary protection containment being the biggest gaps.

Future work should focus on the following: (1) 6-DOF EDL simulation using high fidelity aerodynamic databases and formal design of guidance algorithms, (2) finite-element structural analysis of the modified capsule and MA-ERV interfaces to validate assumed structural mass fractions, (3) validation of quadruped ladder climbing and traversal of Martian terrain during reduced-gravity conditions, (4) formal optimal control solutions for both the low-thrust escape trajectory and the interplanetary return using direct collocation or hybrid differential dynamic programming methods with full ephemeris propagation, (5) planetary protection compliance analysis and design of the containment system details, and (6) formal mass reduction trade study to be completed by closing on Falcon Heavy or finding alternative launchers.

Despite these great challenges, the analyses provided herein show that a singular launch MSR mission is conceptually plausible with projected near-term propulsion, robotics, and ISRU technologies. The value of the study is primarily that it identifies the key design drivers: mass budget closure, structural mass fraction achievement, and planetary protection compliance, that must be resolved before the architecture can be taken from concept to viable mission proposal. The quantitative framework developed in this work is a basis for future trade studies and higher-fidelity analyses intended to resolve these challenges.

References

• 1

  AIAA (2015). Standard mass properties control for space systems. AIAA S-120A-2015.

  • Google Scholar
  • View reference in article

• 2

  ArianeGroup (2018). AESTUS engine data sheet. Germany: Ottobrunn.

  • Google Scholar
  • View reference in article

• 3

  ArmP.ZenklR.BartonP.BeglingerL.DietscheA.FerrazziniL.et al (2019). “SpaceBok: a dynamic legged robot for space exploration,” in 2019 international conference on robotics and automation (ICRA) (Montreal, QC, Canada), 6288–6294.

  • Google Scholar
  • View reference in article

• 4

  BattinR. H. (1999). An introduction to the mathematics and methods of astrodynamics. Reston, VA: AIAA.

  • Google Scholar
  • View reference in article

• 5

  BeatyD. W.GradyM. M.McSweenH. Y.Sefton‐NashE.CarrierB. L.AltieriF.et al (2019). The potential science and engineering value of samples delivered to Earth by Mars sample return. Meteorit. Planet. Sci.54 (S1), S3–S152. 10.1111/maps.13242

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  BerryS. A.LawsC. T.KlebW. L.RhodeM. N.SpellsC.MccreaA. C.et al (2011). “Supersonic retro-propulsion experimental design for computational fluid dynamics model validation,” in 2011 aerospace conference. Big Sky: MT, USA, 1–15. 10.1109/AERO.2011.5747240

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  BettsJ. T. (1998). Survey of numerical methods for trajectory optimization. J. Guid. Control Dyn.21 (2), 193–207. 10.2514/2.4231

  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  BetzA. (1920). Das maximum der theoretisch möglichen ausnutzung des windes durch windmotoren. Z. Gesamte Turbinenwes.26, 307–309.

  • Google Scholar
  • View reference in article

• 9

  BittenR. E.MahrE. M.KelloggR. C. (2013). “Cost estimating of space science missions,” in Aerospace report no. ATR-2013-00108, USA (Washington, DC: NASA).

  • Google Scholar
  • View reference in article

• 10

  Boston Dynamics (2024). Spot specifications. Available online at: https://bostondynamics.com/products/spot/ (Accessed January 05, 2026).

  • Google Scholar
  • View reference in article

• 11

  BraunR. D.ManningR. M. (2007). Mars exploration entry, descent, and landing challenges. J. Spacecr. Rockets44 (2), 310–323. 10.2514/1.25116

  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  CarrierB. L.BeatyD. W.MeyerM. A.BlankJ. G.ChouL.DasSarmaS.et al (2020). Mars extant life: what's next? Conference report. Astrobiology20 (6), 785–814. 10.1089/ast.2020.2237

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  CataldoR. L.BennettG. L. (2011). “US space radioisotope power systems and applications: past, present and future,” in Radioisotopes—applications in physical sciences. Editor SinghN. (Rijeka, Croatia: InTech).

  • Google Scholar
  • View reference in article

• 14

  ChenY.-K.MilosF. S. (1999). Ablation and thermal response program for spacecraft heatshield analysis. J. Spacecr. Rockets36 (3), 475–483. 10.2514/2.3469

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  ConwayB. A. (2010). Spacecraft trajectory optimization (Cambridge, U.K.: Cambridge University Press).

  • Google Scholar
  • View reference in article

• 16

  COSPAR (2021). COSPAR planetary protection policy. Paris, France: COSPAR. Available online at: https://cosparhq.cnes.fr/scientific-structure/panels/panel-on-planetary-protection-ppp/.

  • Google Scholar
  • View reference in article

• 17

  CurtisH. D. (2020). Orbital mechanics for engineering students. 4th ed. Oxford, U.K.: Butterworth-Heinemann.

  • Google Scholar
  • View reference in article

• 18

  EdelbaumT. N. (1961). Propulsion requirements for controllable satellites. ARS J.31 (8), 1079–1089. 10.2514/8.5723

  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  EdquistK.DyakonovA.WrightM.TangC. (2009). “Aerothermodynamic design of the Mars science laboratory heatshield,” in Proc 41st AIAA thermophys Conf (San Antonio, TX: AIAA). Paper AIAA 2009-4075.

  • Google Scholar
  • View reference in article

• 20

  FigueroaO.KearnsS.BollN.ElbelJ. (2023). Report of the independent review board for Mars sample return. Washington, DC: NASA.

  • Google Scholar
  • View reference in article

• 21

  FolknerW. M.WoderC. F.YuanD. N.StandishE. M.PrestonR. A. (1997). Interior structure and seasonal mass redistribution of Mars from radio tracking of Mars pathfinder. Science278 (5344), 1749–1751. 10.1126/science.278.5344.1749

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  FongT.Rochlis ZumbadoJ.CurrieN.MishkinA.AkinD. L. (2013). Space telerobotics unique challenges to human-robot collaboration in space. Rev. Hum. Factors Ergon.9 (1), 6–56. 10.1177/1557234x13510679

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  HoferR. R.GallimoreA. D. (2006). High-specific impulse hall thrusters, part 2: efficiency analysis. J. Propuls. Power22 (4), 732–740. 10.2514/1.15954

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  HoffmanS. J.KaplanD. I. (1997). Human exploration of Mars: the reference mission of the NASA Mars exploration study team, SP-6107. Washington, D.C.: NASA.

  • Google Scholar
  • View reference in article

• 25

  HoffmanJ. A.HechtM. H.RappD.HartvigsenJ. J.SooHooJ. G.AboobakerA. M.et al (2022). Mars oxygen ISRU experiment (MOXIE)—Preparing for human Mars exploration. Sci. Adv.8 (35), eabp8636. 10.1126/sciadv.abp8636

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  Holstein-RathlouC.GunnlaugssonH. P.MerrisonJ. P.BeanK. M.CantorB. A.DavisJ. A.et al (2010). Winds at the phoenix landing site. J. Geophys. Res. Planets115 (E00E18). 10.1029/2009je003411

  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  HumbleR. W.HenryG. N.LarsonW. J. (1995). Space propulsion analysis and design. New York, NY: McGraw-Hill.

  • Google Scholar
  • View reference in article

• 28

  HuntsbergerT.PirjanianP.Trebi-OllennuA.NayarH.AghazarianH.GaninoA.et al (2003). CAMPOUT: a control architecture for tightly coupled coordination of multirobot systems for planetary surface exploration. IEEE Trans. Syst. Man. Cybern. A, Syst. Humans33 (5), 550–559. 10.1109/tsmca.2003.817398

  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  HutterM.GehringC.LauberA.GuntherF.BellicosoC. D.TsounisV.et al (2012). “StarlETH: a compliant quadrupedal robot for fast, efficient, and versatile locomotion,” in Proc. 15th int. Conf. Climbing walking robots (Baltimore, MD), 483–490.

  • Google Scholar

• 30

  HutterM.GehringC.LauberA.GuntherF.BellicosoC. D.TsounisV.et al (2017). ANYmal—Toward legged robots for harsh environments. Adv. Robot.31 (17), 918–931. 10.1080/01691864.2017.1378591

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  HuzelD. K.HuangD. H. (1992). Modern engineering for design of liquid-propellant rocket engines. Reston, VA: AIAA.

  • Google Scholar
  • View reference in article

• 32

  iMARS Working Group (2008). Preliminary planning for an international Mars sample return mission. Rep. Int. Mars Archit. Return Samples Work. Group. Vienna: NASA.

  • Google Scholar
  • View reference in article

• 33

  JustusC. G.JohnsonD. L. (2001). Mars global reference atmospheric model 2001 version. NASA/TM-2001-210961. Alabama: NASA.

  • Google Scholar
  • View reference in article

• 34

  KarczJ. S.DavisS. M.AftosmisM. J.Allen Jr.G. A.BakhtianN. M.DyakonovA. A.et al (2012). “Red dragon: low-cost access to the surface of Mars using commercial capabilities,” in Concepts approaches Mars explor. Houston, TX: NASA.

  • Google Scholar
  • View reference in article

• 35

  LantoineG.RussellR. P. (2012). A hybrid differential dynamic programming algorithm for constrained optimal control problems. Part 1: theory. J. Optim. Theory Appl.154 (2), 382–417. 10.1007/s10957-012-0039-0

  • CrossRef
  • Google Scholar
  • View reference in article

• 36

  LarminieJ.DicksA. (2003). Fuel cell systems explained. 2nd ed. Chichester, U.K.: Wiley.

  • Google Scholar
  • View reference in article

• 37

  LiC.ZhangT.GoldmanD. I. (2013). Terradynamics of legged locomotion on granular media. Science339 (6126), 1408–1412. 10.1126/science.1229163

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 38

  LichterM. D.ViternaL. A. (1999). Performance and feasibility analysis of a wind turbine power system for use on Mars. NASA Tech. Rep. Serv. (NASA). Available online at: https://ntrs.nasa.gov/api/citations/19990081125/downloads/19990081125.pdf (Accessed January 05, 2026).

  • Google Scholar
  • View reference in article

• 39

  LonguskiJ. M.ToddR. E.KonigW. W. (1992). Survey of nongravitational forces and space environmental torques - applied to the Galileo. J. Guid. Control Dyn.15 (3), 545–553. 10.2514/3.20874

  • CrossRef
  • Google Scholar
  • View reference in article

• 40

  MattinglyR.MayL. (2011). “Mars sample return as a campaign,” in Proc. IEEE Aerosp. Conf.,. Big Sky. MT, USA, 1–13.

  • Google Scholar
  • View reference in article

• 41

  MauretteM. (2003). Mars rover autonomous navigation. Aut. Robots14 (2), 199–208. 10.1023/a:1022283719900

  • CrossRef
  • Google Scholar
  • View reference in article

• 42

  McLennanS. M.SephtonM. A.AllenC.AllwoodA. C.BarbieriR.BeatyD. W.et al (2012). Planning for Mars returned sample science: final report of the MSR end-to-end international science analysis group. Astrobiology12 (3), 175–230. 10.1089/ast.2011.0805

  • CrossRef
  • Google Scholar
  • View reference in article

• 43

  MengaliG.QuartaA. A. (2008). Optimal trade studies of interplanetary electric propulsion missions. Acta Astronaut.62 (12), 657–667. 10.1016/j.actaastro.2008.01.037

  • CrossRef
  • Google Scholar
  • View reference in article

• 44

  MinettiA. E.AlexanderR. M. (1997). A theory of metabolic costs for bipedal gaits. J. Theor. Biol.186 (4), 467–476. 10.1006/jtbi.1997.0407

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 45

  MooreH. J.JakoskyB. M. (1989). Viking landing sites, remote-sensing observations, and physical properties of Martian surface materials. Icarus81 (1), 164–184. 10.1016/0019-1035(89)90132-2

  • CrossRef
  • Google Scholar
  • View reference in article

• 46

  MuirheadB. K.NicholasA. K.UmlandJ.SutherlandO.VijendranS. (2020a). Mars sample return campaign concept status. Acta Astronaut.176, 131–138. 10.1016/j.actaastro.2020.06.026

  • CrossRef
  • Google Scholar
  • View reference in article

• 47

  MuirheadB. K.NicholasA.UmlandJ. (2020b). “Mars sample return mission concept status,” in 2020 IEEE aerospace conference. Big Sky. MT, USA, 1–8. 10.1109/AERO47225.2020.9172609

  • CrossRef
  • Google Scholar
  • View reference in article

• 48

  MuscatelloA. C.Santiago-MaldonadoE. (2012). “Mars in situ resource utilization technology evaluation,” in Proc AIAA space conf (Pasadena, CA, USA: AIAA). Paper AIAA 2012-5184.

  • Google Scholar
  • View reference in article

• 49

  NASA (2015). NASA cost estimating handbook. Available online at: https://www.nasa.gov/ocfo/ppc-corner/nasa-cost-estimating-handbook-ceh/ (Accessed January 05, 2026).

  • Google Scholar
  • View reference in article

• 50

  NASA (2016). NASA systems engineering handbook. NASA/SP-2016-6105, Rev. 2. Washington, DC: NASA.

  • Google Scholar
  • View reference in article

• 51

  NASA Glenn Research Center (2013). Advanced stirling radioisotope generator (ASRG). Sandusky, Ohio: NASA Facts. FS-2013-11-011-GRC.

  • Google Scholar
  • View reference in article

• 52

  National Research Council (2011). Vision and voyages for planetary science in the decade 2013–2022. Washington, DC, USA: National Academies Press.

  • Google Scholar
  • View reference in article

• 53

  OnozakiK.YoshimuraH.RossS. D. (2017). Tube dynamics and low energy earth–moon transfers in the 4-body system. Adv. Space Res.60 (10), 2117–2132. 10.1016/j.asr.2017.07.046

  • CrossRef
  • Google Scholar
  • View reference in article

• 54

  OritiS. M. (2015). “Advanced stirling radioisotope generator engineering unit 2 (ASRG EU2) final assembly,” in Proc 12th Int Energy Convers Eng Conf (Cleveland, OH: AIAA). Paper AIAA 2014-3459.

  • Google Scholar
  • View reference in article

• 55

  PetropoulosA. E.LonguskiJ. M. (2004). Shape-based algorithm for the automated design of low-thrust, gravity-assist trajectories. J. Spacecr. Rockets41 (5), 787–796. 10.2514/1.13095

  • CrossRef
  • Google Scholar
  • View reference in article

• 56

  RappD. (2018). Use of extraterrestrial resources for human space missions to Moon or Mars. Berlin, Germany: Springer.

  • Google Scholar
  • View reference in article

• 57

  RaymanM. D.FraschettiT. C.RaymondC. A.RussellC. T. (2006). Dawn: a mission in development for exploration of main belt asteroids Vesta and Ceres. Acta Astronaut.58 (11), 605–616. 10.1016/j.actaastro.2006.01.014

  • CrossRef
  • Google Scholar
  • View reference in article

• 58

  SandersG. B.LarsonW. E. (2013). Progress made in lunar in situ resource utilization under NASA's exploration technology and development program. J. Aerosp. Eng.26 (1), 5–17. 10.1061/(asce)as.1943-5525.0000208

  • CrossRef
  • Google Scholar
  • View reference in article

• 59

  SeedhouseE. (2015). SpaceX’s Dragon: america’s next generation spacecraft. 10.1007/978-3-319-21515-0

  • CrossRef
  • Google Scholar
  • View reference in article

• 60

  SeokS.WangA.ChuahM. Y.HyunD. J.LeeJ.OttenD. M.et al (2015). Design principles for energy-efficient legged locomotion and implementation on the MIT Cheetah robot. IEEE/ASME Trans. Mechatron.20 (3), 1117–1129. 10.1109/tmech.2014.2339013

  • CrossRef
  • Google Scholar
  • View reference in article

• 61

  SinghalS. C.KendallK. (2003). High temperature solid oxide fuel cells: fundamentals, design and applications. Amsterdam, Netherlands: Elsevier.

  • Google Scholar
  • View reference in article

• 62

  SmithD. E.ZuberM. T.FreyH. V.GarvinJ. B.HeadJ. W.MuhlemanD. O.et al (2001). Mars Orbiter Laser Altimeter: experiment summary after the first year of global mapping of Mars. J. Geophys. Res. Atmos.106 (E10), 23689–23722. 10.1029/2000je001364

  • CrossRef
  • Google Scholar
  • View reference in article

• 63

  SpaceX (2026). Falcon heavy. Available online at: https://www.spacex.com/vehicles/falcon-heavy/ (Accessed January 05, 2026).

  • Google Scholar
  • View reference in article

• 64

  SummonsR. E.SessionsA. L.AllwoodA. C.BartonH. A.BeatyD. W.BlakkolbB.et al (2014). Planning considerations related to the organic contamination of Martian samples and implications for the Mars 2020 rover. Astrobiology14 (12), 969–1027. 10.1089/ast.2014.1244

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 65

  SuttonG. P.BiblarzO. (2017). Rocket propulsion elements. 9th ed. Hoboken, NJ, USA: Wiley.

  • Google Scholar

• 66

  SuttonK.GravesR. A. (1971). A general stagnation-point convective-heating equation for arbitrary gas mixtures, R-376. New Jersey: NASA TR.

  • Google Scholar
  • View reference in article

• 67

  TaheriE.JunkinsJ. L. (2018). Generic smoothing for optimal bang-off-bang spacecraft maneuvers. J. Guid. Control Dyn.41 (11), 2470–2475. 10.2514/1.g003604

  • CrossRef
  • Google Scholar
  • View reference in article

• 68

  ThompsonR. W.WilhiteA. W.ReevesD.StanleyD. O.WagnerJ. (2010). Mass growth in space vehicle and exploration architecture development. Acta Astronaut.66 (7-8), 1220–1236. 10.1016/j.actaastro.2009.10.013

  • CrossRef
  • Google Scholar
  • View reference in article

• 69

  TranH. K.JohnsonC. E.RaskyD. J.HuiF. C. L.HsuM.-T.ChenT.et al (1997). Phenolic impregnated carbon ablator (PICA) as thermal protection system for discovery missions, TM-110440. Washington, D.C.: NASA.

  • Google Scholar
  • View reference in article

• 70

  U.S. Congress (2024). Consolidated Appropriations Act. Washington, DC: U.S. Governement. Public Law, 118-47.

  • Google Scholar
  • View reference in article

• 71

  WayD. W.PowellR. W.ChenA.SteltznerA. D.Miguel San MartinA.Daniel BurkhartP.et al(2014). Mars science laboratory entry, descent, and landing system. J. Spacecr. Rockets51(4), 1134–1144.

  • Google Scholar
  • View reference in article

• 72

  WanW.WangC.LiC. L.WeiY. (2020). China's first Mars exploration mission. Nat. Astron4, 721–722. 10.1038/s41550-020-1148-6

  • CrossRef
  • Google Scholar
  • View reference in article

• 73

  WertzJ. R.LarsonW. J. (1999). Space Mission analysis and design. 3rd ed. Hawthorne, CA, USA: Microcosm Press.

  • Google Scholar

• 74

  WertzJ. R.EverettD. F.PuschellJ. J. (2011). Space Mission engineering: the new SMAD. Hawthorne, CA, USA: Microcosm Press.

  • Google Scholar
  • View reference in article

• 75

  WhiffenG. J.SimsJ. A. (2002). “Application of the SDC optimal control algorithm to low-thrust escape and capture trajectory optimization,” in NASA technical report. CA, USA: NASA:

  • Google Scholar
  • View reference in article

• 76

  WongJ. Y. (2008). Theory of ground vehicles. 4th ed. Hoboken, NJ: Wiley.

  • Google Scholar
  • View reference in article

• 77

  YadaT.AbeM.OkadaT.NakatoA.YogataK.MiyazakiA.et al (2021). Preliminary analysis of the Hayabusa2 samples returned from C-type asteroid Ryugu. Nat. Astron.6 (2), 214–220. 10.1038/s41550-021-01550-6

  • CrossRef
  • Google Scholar
  • View reference in article

• 78

  YamC. H.IzzoD.BiscaniF. (2010). “Towards a high fidelity direct transcription method for optimisation of low-thrust trajectories,” in Proc 4th int Conf Astrodyn Tools tech (Madrid, Spain: European Space Agency).

  • Google Scholar
  • View reference in article

• 79

  ZimmermannC. G.NömayrC.KolbM.RuckiA. (2013). A mechanism of solar cell degradation in high intensity, high temperature space missions. Prog. Photovoltaics Res. Appl.21 (4), 420–435. 10.1002/pip.1195

  • CrossRef
  • Google Scholar
  • View reference in article

• 80

  ZouY.ZhuY.BaiY.WangL.JiaY.ShenW.et al (2021). Scientific objectives and payloads of Tianwen-1, China's first Mars exploration mission. Adv. Space Res.67 (2), 812–823. 10.1016/j.asr.2020.11.005

  • CrossRef
  • Google Scholar

• 81

  ZubrinR. (1996). The case for Mars. New York, NY, USA: Free Press.

  • Google Scholar

• 82

  ZubrinR.BakerD.GwynneO. (1991). “Mars direct: a simple, robust, and cost effective architecture for the space exploration initiative,” in Proc. 29th aerosp. Sci. Meeting, Reno, NV, USA. Paper AIAA 91-0328 (Reno: AIAA).

  • Google Scholar

• 83

  ZubrinR. M.MuscatelloA. C.BerggrenM. (2013). Integrated Mars in situ propellant production system. J. Aerosp. Eng.26 (1), 43–56. 10.1061/(asce)as.1943-5525.0000201

  • CrossRef
  • Google Scholar