The ankle-complex joint has an important and clear role during the full gait cycle (Shorter et al., 2013; Conner et al., 2022). At the beginning of the stance phase (initial heel contact), the muscles and tendons around the ankle are responsible for decelerating both the foot rotation before it enters in full contact with the ground, and the forward progression of the body. During mid-stance, the ankle-complex joint helps in maintaining balance and stability as all the user’s weight is solely supported by the ipsilateral limb (while the contralateral limb is in the swing phase). Finally, with the forward tibia progression during stance and the contraction of ankle plantarflexor muscles, energy is stored to be later delivered at the moment of toe-off as plantarflexor torque. This accounts for a large exerted force that is the main source of forward propulsion power during gait (Shorter et al., 2013) (Figure 1).

The role of the ankle-complex joint is compromised in many people with neurological disorders such as stroke, incomplete spinal cord injury or cerebral palsy. This causes a diminished walking ability and the appearance of pathological gait patterns (Beyaert et al., 2015; Armand et al., 2016; Wirz and van Hedel, 2018). The impairments at the ankle level can be related to reduced strength, muscle dysfunction, spasticity, poor selective control or limited range of motion (ROM). Due to the importance of the ankle during gait and in order to limit the consequences of its impaired function, AAFOs are the foremost used type of assistive devices to support walking limitations (Conner et al., 2022; Bayón et al., 2023). For a proper design of AAFO devices, it is key to get a good understanding of the effects of the different pathological gait in the subjects’ walking patterns (Shorter et al., 2013; Conner et al., 2022) (Figure 1).

Weakness in the ankle dorsiflexors (primarily represented by the tibialis anterior) may affect gait in both swing and initial stance phases (Rodda and Graham, 2001; Shorter et al., 2013): during swing, the most common gait deviation is the drop-foot, reflected by an insufficient toe clearance. This deficit is normally compensated by increased knee and hip flexion and hip abduction during swing. Additionally, during initial stance and just after initial heel contact, dorsiflexors’ weakness can also prevent the controlled deceleration of the foot, which is often known as foot-slap.

An impaired function of ankle plantarflexors (with the gastrocnemius and soleus muscles as main contributors) mostly affect the stance phase of gait. Especially, the existence of spasticity or contracture in the gastrocnemius and/or soleus results in equinus pattern (Rodda and Graham, 2001; Brockett and Chapman, 2016), characterized by an abnormal (increased) ankle plantarflexion through most of the stance phase. Additionally, ankle plantarflexors dysfunction also causes a limited push-off power due to the inability of the affected muscles to concentrically contract at the end of the stance phase (Brockett and Chapman, 2016). In general, weakness of ankle platarflexors affects stability, reduces walking speed, shortens step length, and increases the energy cost of walking.

Although the ankle-complex joint has three degrees of freedom (DOF), the primary plane during locomotion and where most of the motion is performed is the sagittal plane (Figure 1). Ideally, AAFOs designed to assist walking in neurological disorders should be able to accommodate dorsi- and plantarflexion assistances depending on the user’s needs. This is translated into different assistive strategies related to the targeted gait impairment: 1) avoiding excessive ankle plantarflexion during swing that causes drop-foot, 2) controlling deceleration of the foot at initial heel contact to inhibit foot-slap, 3) preserving stability during mid-stance while preventing equinus pattern, or 4) generating an assistive plantarflexion torque at the instant of push-off. The last one is also a common strategy used to enhance gait in able-bodied subjects.

The accomplishment of all the above-mentioned actions within a design is a real challenge, especially because adding extra weight to a device located at the ankle (distal joint) could hamper the performance of the user increasing their energy cost of walking (Bayón et al., 2023). Moreover, if the purpose of the AAFO device is to be used during daily-life activities, the design should allow some adaptability to the challenging mobility tasks and ground variations encountered in everyday use (e.g., stairs, ramps, uneven terrains…) (Bayón et al., 2023).

Passive AAFOs are those that do not require any external power supply to generate the assistance. They have the main benefit of being lightweight compared with active AAFOs, so their use is closer to daily basis. To provide assistance, passive AAFOs need to be articulated, and normally they rely on passive elements such as springs, dampers, or elastic bands (conventional non-assistive passive AFOs are out of the scope of this manuscript). In most cases, the design parameters of the passive elements (e.g., the stiffness of the spring) can be optimized only in very specific situations. Thus, these devices have the main limitation of hindering users’ performance for tasks they are not designed for. Regardless of the passive element that is being used, a relevant classification of passive AAFOs can be done based on their working principle and the mechanical configuration of the device (see Supplementary Material).

Passive AAFOs are less adaptable than active (powered) devices due to their simplified design and their inability to control their functions. A trade-off between both solutions are the quasi-passive (or semi-active) devices, where sensors and small motors with low energy consumption are used to modulate the stiffness of passive elements (Supplementary Material). This enhances the adaptability of the devices to changes in gait conditions or environments, allowing the passive element to work in a closed-loop to some extent.

There are certain conditions of motion control and assistance that cannot be accomplished with passive elements due to their limited adaptability and control. Active AAFOs have the advantage of using external power supplies and actuators to address these limitations. Differences between active AAFOs’ designs can be found looking to their mechanical configuration and the control strategies used to assist ankle function (see Supplementary Material).

Adapting to different gait phases, speeds, and terrains is of a great significance for any AAFO. Limiting these adaptations would imply a reduced versatility of the device as well as a hindrance towards the user’s improvement. Moreover, adaptability of the AAFOs does not only refer to adaptations to the environment or walking condition, but also the required adaptation to the user’s progression while using the device.

Passive AAFOs that do not utilize clutches have inherent limitations in terms of adaptability, as the open-loop function of passive elements limits their ability to adapt to different walking conditions and tasks. These passive devices are not capable of providing assistance to either dorsi- or plantarflexion without hindering the other. The problem is resolved to some extent by using body-weight dependent clutches [e.g., (Yandell et al., 2019; van Hoorn et al., 2022)]. These types of clutches can detect the switch between different phases of gait regardless of the speed of the user, but their effectiveness has not been tested yet when walking over different terrains.

Quasi-passive AAFOs can more accurately switch the function of the passive elements to better react to changes in gait phases (Wang C. et al., 2022). Some of them can adjust the stiffness of the AAFO joint to better match the stiffness of the biological ankle (Allen et al., 2021), or directly adjust the assistance provided by the device on the subject (Diller et al., 2016; Oba et al., 2019; Allen et al., 2021).

Finally, active AAFOs are able to control dorsi- and plantarflexion separately, thereby, they offer more possibilities for adaptation irrespective of whether their main purpose is to support drop-foot [e.g., (Blaya and Herr, 2004; Awad et al., 2017b; Kim et al., 2020)], foot-slap [e.g., (Kim et al., 2011; Choi et al., 2018; Chen B. et al., 2022)], push-off [e.g., (Ward et al., 2011; Choi et al., 2018; Acosta-Sojo and Stirling, 2022)] or full gait [e.g., (Lerner et al., 2018; Siviy et al., 2020; de Miguel-Fernandez et al., 2022b)]. In fact, apart of being able to adapt to walking velocities, they also have the ability to control their actuator to be active during some phases of the gait and transparent during other phases (Bayón et al., 2022).

Regarding the subjects involved in the experiments, three main validation strategies have been reported across the reviewed articles (Figure 4). Some authors just reported technical validations regarding the proper functioning of the device and the assistance generated in terms of force and torque [e.g., (Diller et al., 2016; Choi et al., 2018; Leclair et al., 2018)] but without directly testing in subjects. When subjects were involved in the experiment, we can distinguish between those that involved healthy [e.g., (Malcolm et al., 2018; Oba et al., 2019; Wang et al., 2020)] or impaired subjects [e.g., (Sekiguchi et al., 2020; Fang et al., 2022)]. Remarkably, the AAFOs considered in this manuscript are more extensively validated with healthy than with impaired walking subjects. This may be due to two main reasons: 1) some AAFOs are directly developed to empower healthy subjects and increase their endurance while walking; 2) healthy subjects are considered as a first step for the validation of AAFOs, even if they are designed to assist paretic patients.

When reporting validations with subjects, authors focused on the assessment of three main aspects: 1) the lower-limbs kinematics, related to functional aspects of gait; 2) the muscular adaptations that occur when users received the assistance provided by the AAFO; and 3) the metabolic cost of walking.

AAFOs are always intended to assist ankle motion based on the specific user’s needs, which is done by the provision of assistive ankle torques. The main reason to select an AAFO design over others is the ability of the device to tailor the assistance to address those specific needs.

The principal design-related benefit of passive and quasi-passive devices compared with active ones is their reduced weight, as they do not use heavy actuators or batteries. From these, ADR-AAFOs are the passive devices closer to the market. They support the ankle during the entire gait cycle, providing bidirectional torques by means of passive springs (Ultraflex Systems, 2009; Becker Orthopedic, 2018; Ottobock, 2018; Fior and Gentz, 2021). However, these devices are reported to be still bulky, and it is difficult to adjust their spring stiffness to the specific user’s pathological gait (Kerkum et al., 2014; Kobayashi et al., 2017). Moreover, the fact of providing a constant stiffness along the whole gait cycle without a dedicated timing for the storage and release of energy, may hinder the subject’s residual capabilities. This, together with the lack of evidence that proves their ability to adapt to different scenarios and mobility tasks encountered in everyday life, can be considered as the big limitations for these solutions (Kerkum et al., 2015; Firouzeh et al., 2021; Bayón et al., 2023).

Spring-cam passive designs (Rodriguez et al., 2018; Sekiguchi et al., 2020) tried to introduce a concrete benefit over the ADRs: the possibility of customizing a non-linear torque-angle curve depending on the user’s needs. This advantage allows the adoption of alternative tailored designs to counteract for diverse pathological gaits and the allowance of some adaptation to different environments and settings. Although still lighter than active solutions, these devices generally are a bit heavier than other passive ones like the ADRs.

With the appearance of passive clutch-dependent devices (Hirai et al., 2006; Collins et al., 2015; Leclair et al., 2018; Pardoel and Doumit, 2019; Yandell et al., 2019; Wang et al., 2020; Liu et al., 2021; van Hoorn et al., 2022) researchers made the next effort to gain more control over the timing of the delivered assistance within the gait cycle. This potential benefit has been further exploited by quasi-passive AAFOs with the aim of reaching higher torque levels and kinematic compatibility than with passive ones (Zhang et al., 2015; Dezman et al., 2018; Kumar et al., 2020; Wang C. et al., 2022). The shortcoming for clutch-based solutions is that normally, they provide assistance only on a single direction (plantarflexion or dorsiflexion).

Still, (quasi) passive devices in general have not been extensively tested to prove their good performance in adapting to different users and walking terrains. On the contrary, active AAFOs have the major benefit of being able to provide large levels of assistance fully controllable in amplitude and timing, which enables them to adapt to different velocities or varying ground conditions (Shorter et al., 2011; Yeung et al., 2017; Lerner et al., 2018; Huo et al., 2019; Meijneke et al., 2021; Chen J. et al., 2022). However, their application is restricted to the performance of their controllers and actuators. For example, most of the devices that have a rigid transmission use position control rather than torque control. These devices are able to precisely control the final kinematics of the device, and therefore of the joint, but without considering the exerted torque on the subject, which can be dangerous if it is not properly controlled. They are mainly applied to patients who do not have control over their limbs and who are not able to oppose the movement (Choi et al., 2018; Weerasingha et al., 2018; Huo et al., 2019; Lee et al., 2021). To alleviate this strict position tracking, some researchers used impedance control to allow some deviation from predefined trajectories (Lee et al., 2021) or by using low-transmission brushless DC motors, which have low back-drivability, allowing users to have safer interaction with the AAFO (de Miguel-Fernandez et al., 2022a; de Miguel-Fernandez et al., 2022b; Nabipour and Moosavian, 2021).

The stiffer the actuator, the more delicate and accurate the controller should be to track the user’s motion or intention. Therefore, most of the active devices use compliant actuation that decreases the demands of the controllers due to the inherent adjustable behavior of the actuator. Starting from the geared actuator, the straightforward compliant system is formed by adding an elastic element, i.e., a spring, in series with the actuator (Kim et al., 2011; van Dijk et al., 2017; Durandau et al., 2022). These SEA actuators reduce the restriction imposed by the low back-drivability of the rigid transmission, being a compromise solution between maximal torque and weight that can fit well to assist impaired patients. However, depending on the device implementation, their size and bulkiness could not be yet suitable for daily use, so improving these is an open challenge.

An important issue about the assistance provided by active AAFOs is the strategy followed to deliver a timed assistance without disturbing the subject. In this sense, non-compliant devices with a rigid power transmission can only rely on control strategies that synchronize the subject’s gait with the action of the AAFO. However, the nature of compliant mechanisms allows for a slight absorption of the hampering effect of imprecise controllers, which permits the transparent behaviour of the AAFO when the assistive torque is not given.

Devices based on PAMs are characterized by the best compromise between exerted force and lightweight (Kinnaird and Ferris, 2009; Koller et al., 2015; Galle et al., 2017). However, their slow dynamics only allow their use in subjects walking or moving at a slow velocity, i.e., impaired or elderly subjects. In addition, these devices are limited by their necessity of a compressed air supply to power them, which constrains their application to assist daily life activities.

Finally, AAFOs driven by cables have the potential of being the most lightweight active approach (Bae et al., 2018a; McCain et al., 2019; Slade et al., 2022). By placing the actuator in a proximal location (mostly around the pelvis), the extra inertia added to the distal segment of the limb is reduced. This feature would enable them to be used in daily life scenarios, although they have not been extensively tested in such conditions. As a drawback, cable-driven AAFOs are only able to assist the motion in the traction direction of the cable, normally the plantarflexion direction of the ankle. This would prevent the assistance during the full gait cycle, but could be enough to assist certain gait functions such as push-off. Alternatively, by adding more than one cable and actuator, the device could assist both dorsi- and plantarflexion (Rewalk Robotics, 2023).

By analyzing the experimental validations performed with the reviewed AAFOs (Figure 4), we can conclude that active devices have aroused more interest in the scientific community, dealing with a greater number of related articles. Moreover, none of the reviewed quasi-passive AAFOs have been finally validated with impaired subjects, even when they were designed to do so. This fact may imply that the readiness-level of quasi-passive solutions is lower than for purely passive or active devices and, therefore, further research effort should be done before this technology can be used in real-life scenarios by people with impairments.

The push-off assistance is the preferred goal for all three AAFOs categories (i.e., passive, quasi-passive, and active). In most cases, the target populations for this type of support are healthy (empowerment), or aging and people who suffer from muscle weaknesses (assistance). Most of these devices have only been validated with experiments on healthy subjects [except (Sekiguchi et al., 2020) for passive, and (Ward et al., 2011; Takahashi et al., 2015; Awad et al., 2017a; McCain et al., 2019; Tamburella et al., 2020; Gomez-Vargas et al., 2021; Harvey et al., 2021) for active]. A reason for the limited evaluation might be that most of these devices require subjects with good balance capabilities as the devices do not normally account for that. For future use of these devices in populations with impairments it is key that they account for balance loss, assisting the user in that regard, or at least not being detrimental.

In the case of passive and quasi-passive devices, the capacity to support push-off is normally assessed by reductions in muscular activity. All the reviewed (quasi) passive studies reported reductions of plantarflexors muscular activity. However, dorsiflexors like the tibialis anterior, were normally not assessed or even showed an increase of activity (Collins et al., 2015; Wang C. et al., 2022). Only (Kumar et al., 2020) reported a reduction in tibialis anterior and soleus at the same time, but their validation was only performed on a single healthy subject. These results suggest that extensive validations are still required for these devices.

Regarding active AAFOs, assisting push-off is the preferred objective for most cable-driven and PAM configurations. In this case, the validations focused on reducing the metabolic cost of walking in healthy subjects and improving kinematics and kinetics for paretic ankle propulsion. Based on these metrics, both approaches appeared to be valid alternatives that achieved their assistive purpose.

After push-off, the most common assistance is full-gait. For passive devices, ADRs and spring-cam are the approaches that allow for a double dorsi- and plantarflexion support (Kerkum et al., 2014; Wren et al., 2015; Kobayashi et al., 2017) but have their limitations. As mentioned before, the assistance provided is not timed within the gait cycle. This implies that instead of assisting the residual capabilities of the subject to perform a specific ankle motion, they provide stability while might restrict motion. Most of the ADR-AAFOs were tested with impaired subjects and some qualitative improvements in subjects’ kinematics were reported. Nevertheless, no quantitative extensive discussion over their effects on the users is published (e.g., to evaluate if the added mass of the device to provoke those improvements is justified). Therefore, in spite of these promising results, there is not solid evidence of their benefits when assisting the complete gait cycle.

Active AAFOs offer more possibilities to assist full gait but for some mechanical designs, this comes at the cost of a double configuration. Only for the usually employed SEAs or rigid transmission the actuation provided in both dorsi- and plantarflexion is intrinsic to their design. However, PAM-based or cable-driven devices could also allow for that following a double configuration, with the actuator placed in parallel to both the agonist and antagonist muscle groups. In spite of this potential solution that would extend the capabilities of these groups of devices, the double configuration has only been explored by cable-driven AAFOs [e.g., (Awad et al., 2017b; Gomez-Vargas et al., 2021; Rewalk Robotics, 2023)].

As mentioned earlier, due to the lack of consensus in the used experimental procedures between studies, it is very difficult to find common reported criteria for an objective comparison of the performance of different AAFOs.

When addressing the functional performance of their device, technological indicators that describe the physical capability of the AAFO are normally provided. In that regard, researchers tried to report results on, for example, joint kinematics, spatio-temporal parameters, interaction forces, adaptability to walking speed, and effects on muscular activity or energy consumption. However, not all these topics were addressed for all devices, and even more importantly, the topics contemplated were assessed following different methodologies or with different target populations. Therefore and in order to have a fair quantitative comparison, it is essential to make sets of bench-marking criteria that researchers can follow to evaluate their devices, which is the approach also proposed by Torricelli et al. (2020).

Additionally, there is also a lack of evaluations of the devices in diverse conditions that are decisive for representing daily-life situations. Currently, devices are mostly evaluated during flat overground walking, treadmill, or just technically. Experiments on non-standardized settings like uneven terrains, when maneuvering around, or just start and stop gait for instance, have not been sufficiently discussed. Thus, although it is possible to hypothesize, for example, that AAFOs with body-weight dependent clutches are superior to spring-cam or angle-dependent clutches in providing assistance while walking on a slope, no evidence is available for proving this claim.

The effect of the AAFOs on users’ balance or user’s safety further than locomotion support are important aspects that are scarcely considered in the literature. There are some studies with active AAFOs that investigated the capacity of the device on improving balance capabilities of the users (Bayón et al., 2022; Gonzalez et al., 2022), but they were not focused on assessing the mechanical configuration of the orthosis itself, but on evaluating a specific controller [except (Harvey et al., 2021)].

Finally, the users’ perceived experiences in terms of perceptual, emotional and cognitive aspects are also barely reported. A reason for that might be the small number of studies and evaluations with human beings at long term, which limits the inclusion of questionnaires or human-in-the-loop experiments with variations of the technical device that report experiences of final-users. These reports would be very useful, for example, to assess the perceived comfort, control (embodiment and agency) and required effort when using an specific AAFO.

Based on the above-mentioned points, the advantages and disadvantages reported in Section 6.1, and the data found on the literature (see Supplementary Table S2 in the Supplementary Material), we provide a subjective analysis of the clusters created in this article in terms of the readiness level we think they have in terms of functional performance and user experience (Tables 2, 3 respectively). The features included in these tables are categorized using the framework of (Torricelli et al., 2020), and can also be used as a potential set of criteria that researchers can utilize when developing new devices in order to ensure a future consistent evaluation.

There have been quite some advances in passive AAFOs in the last years, but still these devices are not robustly prepared to address all kinds of situations encountered in everyday life. Started from less than 10 years ago, quasi-passive solutions represent a compromise between passive and active devices, resulting in an increasing trend that opted for using low electrical power to activate clutches. A trend for these devices (started in 2016) was to use other types of actuation rather than just springs and dampers for applying the assistance. The electroadhesive and dielectric elastomer actuators are examples of this. However, most of the passive and quasi-passive approaches have the limitation of providing assistance only towards dorsi- or plantarflexion, but not both. A groundbreaking step for (quasi) passive designs would be to have a device able to provide a configurable assistance on both directions with minor modifications.

Active AAFOs have the potential of being fully controllable to assist several gait functions with a single design. These devices are now evolving to lighter and more compliant solutions that should be transparent for the user while they are not providing assistance. Following the tendency of passive devices to assist activities of daily living, active AAFOs should also be tested in unrestricted environments to assess their performance when assisting gait under more challenging and rapidly varying conditions (variations of terrain and gait velocity).