A newer HIIT-related variation is high intensity functional training (HIFT) which is a combination of functional multi-joint movements. These movements are adjustable to any fitness level and elicit greater muscle recruitment than more traditional exercises (Feito et al., 2018). These functional training elements, i.e., exercises that mimic movements of daily living (e.g., squats and lunges), have been shown to simultaneously improve strength and balance (Weiss et al., 2010). While HIIT exercise is characterized by relatively short bursts of repeated vigorous activity, interspersed by periods of rest or low-intensity exercise for recovery, HIFT utilizes constantly varied functional exercises and various activity durations that may or may not incorporate rest (Feito et al., 2018). The commonly practiced combination of both approaches is fHIIT.

A recent study compared effects of moderate aerobic exercise and circuit-based fHIIT on motor performance and exercise motivation in untrained adults (Wilke et al., 2019). The circuit-based fHIIT enhanced physical functions (strength and endurance) and motivation to exercise more effectively than the moderate condition. Another study examined the physiological effects of an fHIIT program on endurance and strength of physically active adults over a 4-week period and found rapid physiological improvements in strength as well as in aerobic and anaerobic capacity (Kliszczewicz et al., 2019). fHIIT seems to be a beneficial variation of HIIT as its protocols allow for multiple performance and physiological adaptations that are not observed by training using unimodal HIIT methodology (Feito et al., 2018). fHIIT combines the best of HIIT and HIFT, benefits the whole body (endurance, strength, coordination, flexibility, etc.) and transfers more to daily life activities (McRae et al., 2012; Buckley et al., 2015; Feito et al., 2018; Menz et al., 2019).

Today's fitness market is reacting to this and provides special fHIIT classes with different foci (e.g., BodyAttack®) which—similar to spinning classes—enable an intense and socially motivating group workout on a holistic level. Mobile fitness apps such as Freeletics¹ further provide options for digital fHIIT-like training for users “on the go” and allow them to share, compete, and cooperate with one another. Although fitness providers frame HIIT and fHIIT as motivating as possible, it remains an extremely challenging training approach.

In today's digital age, exergames (Oh and Yang, 2010)—games that are controlled by physical exercises and provide an additional cognitive challenge for the player—are being explored as a suitable tool to introduce more people to effective training approaches and motivate them to keep on track.

Studies on exergame training in different target populations such as older adults, children, adolescents or patients indicate effects on cognitive (e.g., executive function, attention, and visual-spatial skills) (Li et al., 2016; Joronen et al., 2017; Lee et al., 2017; Byrne and Kim, 2019; Kappen et al., 2019; Stojan and Voelcker-Rehage, 2019), physical (e.g., energy expenditure, HR, and physical activity) (Lu et al., 2013; Lee et al., 2017; Byrne and Kim, 2019; Tondello et al., 2019), and mental (e.g., social interaction, self-esteem, motivation, and mood) (Macvean and Robertson, 2013; Lyons, 2015; Lee et al., 2017; Martin-Niedecken and Götz, 2017; Valenzuela et al., 2018) aspects. Generally, exergames are well-known for their playful combination of physically and cognitively challenging tasks and thus provide dual-domain training which promises greater effects compared to traditional single-task training approaches (Huang et al., 2014; Hardy et al., 2015; Benzing and Schmidt, 2018; Kappen et al., 2019).

Besides specific effects of exergame training, it is further known for its appealing and motivating impact, especially in physically inactive populations (Wüest et al., 2014; Hoffmann et al., 2016). By providing different players [of different motivational types (Isbister, 2016)] with audio-visually, narratively appealing, and immersive game scenarios, exergames shift players' (cognitive) focus onto the playful experience. This makes it easier to engage with a physically challenging workout (Xiong et al., 2019). Therefore, exergames have successfully been shown to increase training adherence (Kajastila and Hämäläinen, 2015), long-term motivation (Márquez Segura et al., 2013), engagement (Mueller and Isbister, 2014), immersion (Wüest et al., 2014), and flow (Sinclair et al., 2007) in players from different populations.

In the context of cHIIT, only few studies exists that investigated feasibility of exergames specifically designed for cHIIT with regards to physiological training outcomes or qualitative factors such as motivation and enjoyment. To the best of our knowledge, no exergames have been specifically designed and evaluated for fHIIT as of yet.

de Bruin et al. (2019) investigated the feasibility and effects on cardiovascular fitness of an exergame-based HIIT program in untrained elderly people. The 4-week training included a cognitively-simple game which required fast steps for the intense training phases, and games that were cognitively more but physically less challenging for the low-intensity phases. In the low-intensity phase, participants ranged within 50–70% of their maximum heart rate (HR_max). Both used a step-based platform as game controller. They found that the exergame-based HIIT was a feasible and well-accepted approach and led to the intended physical intensity (70–90% of HR_max). Furthermore, their collected qualitative feedback identified certain aspects which could increase study outcomes in future iterations (e.g., game music, more audiovisual feedback, and increased challenge).

Farrow et al. (2019) compared different in-game conditions (allowing participants to race against their own performance or by increasing the resistance) in a head-mounted virtual reality (VR) exergame-based HIIT on an ergometer against traditional ergometer-based HIIT in physically inactive young adults. They found that VR exergaming increased enjoyment during a single bout of HIIT and led to an average of 74–89% of HR_max over all tested conditions in untrained individuals. Furthermore, the presence of a virtual ghost to compete with appeared to be an effective method to increase exercise intensity of VR-based HIIT.

Barathi et al. (2018) proposed and evaluated an interactive feedforward approach (a method based on competition with oneself, i.e., against an improved self-model of the player) to rapidly improve performance in a HIIT cycling VR exergame. They found that the interactive feedforward method led to improved performance (participants' average HR was above 80% of HR_max) while maintaining intrinsic motivation and was superior to competing against a virtual competitor.

A different VR-HIIT exergame was developed for a rowing machine by Keesing et al. (2019). They utilized gameplay mechanics and the synchronization of rowing rhythm with music rhythm to automatically induce HIIT without the need for a physical instructor. They reported that gameplay and music were both effective at inducing HIIT, but music had a stronger effect on both performance and enjoyment.

Haller et al. (2019) investigated the effects of virtual spectators (and their rhythmic clapping based on participants' ergometer speed) on motivation during a HIIT-exergame. They found that virtual crowd feedback increased cycling speed and participants' HR [to around 171 beats per minute (bpm); percentages of HR_max were not reported].

Finally, Moholdt et al. (2017) compared HIIT with an online multiplayer ergometer-based exergame to walking in male students. Their exergame elicited an average intensity of 73–83% of HR_max and a higher enjoyment than walking.

In summary, the evaluated exergames did not feature full-body functional exercises, nor necessarily a comprehensive, meaningful, audio-visually appealing, and adaptive game design. By this, we mean that—besides different training approaches—these exergames did not follow a holistic design approach covering all design levels of an exergame (Martin-Niedecken and Mekler, 2018; Martin-Niedecken et al., 2019) as well as taking into account potential interdependencies and interaction effects between these, which can affect the targeted game experience. An attractive and effective exergame design encloses the player's moving and sensing body (Bianchi-Berthouze et al., 2007; Mueller et al., 2011) and allows for effective and playful exercises (Marshall et al., 2016). These exercises in turn are mediated by the game controller technology which should be easily and naturally embedded into the moving player's body scheme (Pasch et al., 2009; Kim et al., 2014; Shafer et al., 2014). Moreover, the virtual game scenario represents the player's bodily input in the virtual environment and provides audio-visual as well as haptic or tactile feedback for the player and their reacting body (Shaw et al., 2017). Furthermore, the aforementioned exergames did not necessarily feature individually adjustable cognitive and physical challenges (Sinclair et al., 2009). Thus, to the best of our knowledge, while there are HIIT exergames, there are no exergames specifically designed for fHIIT, nor studies investigating them.

The ExerCube (Martin-Niedecken and Mekler, 2018; Martin-Niedecken et al., 2019) is an immersive mixed-reality fitness game. Players are surrounded by three walls, which serve as projection screens and a haptic interface for energetic bodily interactions. A customized motion tracking system tracks players' movements via HTC Vive trackers (attached to their wrists). To ensure an ideal workout experience [in terms of attractive design and effective exercises (Sinclair et al., 2009)] for a wide spectrum of players with different skill sets, the ExerCube continuously adapts game difficulty to players' individual fitness and cognitive skills. Training intensity is measured via continuous HR tracking (i.e., players wear a HR-sensor chest strap) and set to an individual pre-defined HR training range. Cognitive skills are measured via in-game performance (reacting to visual stimuli at the right time).

The Sphery Racer (see Figure 1) is a single-player game experience designed for the ExerCube setting. It was developed in several iterations based on the prototype presented and evaluated by Martin-Niedecken and Mekler (2018) and Martin-Niedecken et al. (2019), and it is now being employed as a research object by several research groups. In collaboration with the ExerCube's development team, we modified the game design (game mechanics and audio-visual design), the level structure and the HR-based game adaption algorithm to be comparable with a cfHIIT.

Like its prototypical predecessor, Sphery Racer asks players to progress along a fast-paced race track via an avatar on a hoverboard. The motion tracking system transfers player movements (based on a functional workout) onto this avatar and thus on the virtual racing track. Along the race, players are challenged by obstacles that require physical exercises (e.g., squats, lunges, and burpees) and by an additional cognitive challenge including quick information processing, which exercise has to be performed when (i.e., reaction, time, and coordination challenges).

The game starts with an on-boarding tutorial scene during which the game is calibrated to the exact height of the player. After successful calibration, the player's avatar drives onto the racing track and to the first instructional pitstop sequence (see Figure 2). The game contains five training pitstops (~0.5–2 min. each, see Table 1), which serve as tutorials to become familiar with the respective steering movements. All exercises are instructed audio-visually (i.e., the avatar shows the exercise, written instructions are added, and a voice provides additional hints). The exercises start with low-to-moderate intensity (in terms of both physical and cognitive demand) and over time gradually increase until reaching high-intensity exercises (e.g., skippings and burpees).

After each pitstop, players return to the racing track, where they perform all thus-far learned movements in five racing sections. To integrate a warm-up phase followed by gradually intensifying level design, the racing section durations range from 2.5 min (first and second sections), to 5 min (third and fourth), and finally 10 min (last section). A complete workout session in the ExerCube takes 26–28 min.

The physical and cognitive game difficulty adjustments are gradually adapted independently over all training levels on a 10-point difficulty scale, where one level is defined as one step on the 10-point scale (e.g., from 5 to 6). Like previously, the algorithm determines players' individual calculated HR_max (CHR_max) based on the following formula (Nes et al., 2013): (1)CHRmax=211-0.64×age A comprehensive fitness study proved that this formula adequately explained HRmax by considering an age range of 19–89 years (Nes et al., 2013). Previously the ExerCube's algorithm also aimed toward reaching a high intensity training level (80–90% of HR_max), by a less finely tuned algorithm (HR <150 bpm for 0.5 min: increase speed slightly by one level; HR >175 bpm for 1 min: decrease speed slightly by one level; HR >190 bpm: decrease speed strongly by two level). However, the algorithm was not found to reach this training intensity. For the purpose of the presented study, we refined the algorithm: During the racing sections, the game aimed to get players to a specific HR range (70–90% of CHR_max) and then kept them at this level. Outside of this range, a lower HR lead to an increase in physical challenge, i.e., speed, exercise frequency (one level per check), while a higher HR lead to a decrease (once 100% of CHR_max was reached, this decrease was sped up by three levels to ensure players' safety). For the first two racing sections, the system employed a strategy for increasing players' HR, i.e., when 70% of CHR_max has not been reached, it checked actual HR_max every 30 s (whereas every 60 s otherwise). For the subsequent racing sections (3–5), the system checked more often (every 20 s in the increasing phase, and 10 s when above 90% of CHR_max).

Since the focus of the presented study was to compare the physical training intensity, we employed the same algorithm for cognitive game difficulty adjustment as used by Martin-Niedecken et al. (2019). The main cognitive challenge of the game related to how early players were visually instructed about the direction (right or left) of the upcoming exercise (e.g., a yellow gate rotates to the right for a high touch). If a player performs error-free for 20 s, the cognitive difficulty increased by one level (resulting in a delayed display of the direction of the next exercise) until they made three mistakes within 20 s, inducing a difficulty decrease by one level (resulting in an earlier display of the direction of the next exercise).

However, since the first ExerCube iteration (Martin-Niedecken et al., 2019), a new physical-cognitive challenge was added to the game scenario: players are rewarded with up to three stars depending on their timing. The audio design was developed further to emphasize the background music's rhythm, emphasize feedback via sound effects, and incorporate audio instructions. Finally, the visual feedback system was iterated for clarity.

To provide a comparable training protocol, we created a specific cfHIIT (see Figure 1) that was as close to actually practiced fHIIT as possible, still comparable to the ExerCube's training protocol. The ExerCube's exercises and intervals thus served as a basis to ensure a similar physical load in both conditions (Table 1).

The fHIIT protocol consists of five blocks. It started with a short warm-up block (block 1: 5 min stretching and toning) followed by four interval training blocks, whereby blocks 2–4 included two different exercises (e.g., jump squats and lunge jumps) and block 5 included three different exercises (e.g., skipping). This ensured an increase in physical load toward the end of the training session and matched the ExerCube's last interval. Each (non-warm-up) exercise interval consisted of 20 s of workout (alternately performing the respective exercises) and 10 s of rest. These 30 s workout-rest phases were repeated 8 times (blocks 2–4) or 12 times (block 5), leading to a total duration of 4 min (blocks 2–4) or 6 min (block 5). Interval blocks were separated by short breaks of 1.5 min (blocks 2–4) or 2 min (between blocks 4 and 5). Overall, the cfHIIT lasted about 28 min.

Participants were instructed by the coach that they could individually adapt exercise intensity themselves by choosing a lower or higher level of the initial exercise (e.g., lunges instead of lung jumps) based on their subjective experience of their physical exertion. Additionally, participants were offered a mobile device positioned on the floor nearby the participants that showed their current HR in real time to allow them to keep track of their individual training intensity.

The cfHIIT session was accompanied by music that also functioned as a timer for the intervals. The selected music was specifically composed for HIIT with a pace of 128 beats per minute (bpm), while slower songs were chosen for the breaks in-between intervals to support recovery. The music was played to enhance participants' motivation, to facilitate similar conditions as the ExerCube training (accompanied by a specifically developed and adaptive sound design), and to present a realistic scenario.

Two study objectives were determined to investigate the manifestation of objective and subjective components of a single fHIIT exergame session in comparison to a single cfHIIT session. The primary objective of this study was to evaluate the objectively and subjectively experienced training intensity of a single fHIIT exergame session in comparison to a single cfHIIT session. The secondary objective was to assess motivation, flow, and enjoyment during the two training approaches.

The comparative study was set up as a within-subjects design allowing the comparison of two different training methods: an ExerCube vs. cfHIIT session. Whereas, the ExerCube session was performed as a single-player session and controlled by a certified fitness coach, the cfHIIT was performed in small groups of 2–3 participants and instructed by the same coach. Although the ExerCube was mainly self-explanatory, the coach instructed participants and supervised them throughout the session. In the cfHIIT condition, the coach directly instructed the exercises and performed the training session together with the participants. In both sessions, the coach provided corrections, verbal support and cheers, if needed. Moreover, participants did not interact (physically or verbally) with each other in the cfHIIT session.

The sample size was calculated a priori based on a previous study comparing two HIIT protocols regarding cardiac responses (Schaun and Del Vecchio, 2018). Their study showed the following during exercise: average HR (HR_avg) 144.2 ± 11.9 bpm and 130.6 ± 10.4 bpm. Considering an 80% power and 5% significance level, a sample size of 11 subjects would have been necessary. To account for a potentially smaller difference in HR_avg between the two training types of our study and regarding possible losses or refusals, the final sample size was set to 16–20 participants.

Twenty participants (10 male, 10 female) were recruited by word-of-mouth and by emailing, without offering any financial compensation for the attendance. The selected study population included healthy young adults (self-reported using a health questionnaire) aged 18–35 years (M = 23.8 years, SD = 3.2). Fifteen participants had experience with exergames, five did not. Participants were excluded from the study if one of the following exclusion criteria was met: (1) history of cardiovascular issues or musculoskeletal injuries that would prevent training participation, (2) asthma (not controllable), (3) pain that would be reinforced by sportive activities, (4) pregnancy.

The recruited participants reported an average exercising time of M = 300.3 min/week (SD = 167), and reported their subjective fitness as an average of M = 3.9 (SD = 0.9) on a 6-point scale (1 = poor, 2 = satisfactory, 3 = average, 4 = good, 5 = very good, 6 = competitive). Their resting HR measured M = 70.3 bpm (SD = 9.9)—we thus calculated their CHR_max at M = 195.8 (SD = 2.0).

We distinguish between primary outcomes—relating to training intensity—and secondary outcomes, which relate to the qualitative experience of the two training types.

After study explanation, each participant gave written informed consent. Afterwards, participants filled out a demographic questionnaire to screen for inclusion and exclusion criteria and to assess baseline characteristics such as gender, age, physical activity time, fitness status, and exergame experience. Participants were randomly assigned to one of the two trainings. Each training session lasted 26–28 min. After the training, participants rated their perceived physical and cognitive exertion using the Borg Scale and answered questionnaires covering motivation, flow, and enjoyment. A training session with subsequent questionnaires was then repeated with the other type of training on a different day (but same time of day), after a minimum of 4 days and a maximum of 14 days in-between. The study procedure is illustrated in Figure 3.

Statistical analysis was conducted in SPSS (IBM SPSS 26). The level of significance was set at p < 0.05. The comparison of the data was performed using Wilcoxon signed-rank tests. Wilcoxon signed-rank was used because the assumptions for parametric statistics were not fulfilled. Correlations were calculated using the Spearman correlation coefficient. See Cohen (2013) for an overview of thresholds for correlation coefficients and effect sizes.

Table 2 presents the results from the comparison of the average and maximal measured HR and Borg values between the ExerCube and cfHIIT sessions. Absolute (z = −2.878, p = 0.003, r = 0.46) and relative (z = −2.837 p = 0.005, r = 0.45) average HR values were significantly higher for the cfHIIT session than for the ExerCube training. For the maximal HR, no significant differences were found for absolute (z = −0.262, p = 0.806, r = 0.04) and relative (z = −0.302, p = 0.388, r = 0.05) values. In terms of Borg values, the cfHIIT resulted in a significant higher physical Borg rating (z = −3.020, p = 0.001, r = 0.48) than the ExerCube session. No significant difference was measured for the cognitive Borg (z = −1.603, p = 0.113, r = 0.25).

The questionnaire data showed significant differences for intrinsic motivation (z = −3.566, p < 0.001, r = 0.56), overall flow score (z = −3.663, p < 0.001, r = 0.58), absorption by activity (z = −3.436, p = 0.001, r = 0.54), perceived importance (z = −2.518, p = 0.012, r = 0.40), and physical activity enjoyment (z = −3.884, p < 0.001, r = 0.61), see Table 3. For all of these factors, scores were higher for the ExerCube training session. Additionally, a significant correlation (r_s = 0.365, p = 0.021) was found between average HR and physical Borg values across all training session data (ExerCube and cfHIIT). No significant correlations were found for Borg_physical-HR_max (r_s = 0.276, p = 0.084), Borg_cognitive-HR_avg (r_s =-0.224, p = 0.164), or Borgcognitive-HRmax (rs = −0.133, p = 0.412).

The primary interest of the presented work was to objectively and subjectively investigate the intensity of an exergame-based fHIIT with the ExerCube, and thus to explore the feasibility of a specifically designed exergame as a suitable training tool for effective fHIIT. Besides this general proof of feasibility (i.e., reaching the 70–90% range of CHR_max), we found implications which seem to be important for future research and development work.

Additionally of interest to our work was the comparison of the subjective training experience of exergame-based fHIIT with the ExerCube to a cfHIIT. Besides the general proof of feasibility of the ExerCube to be an attractive fHIIT exergame, we again found implications for future research and design toward more appealing exergames.

One limitation of the study consists of the differences between our two training stimuli. While we endeavored to design the two conditions to be as comparable as possible, we also wanted to keep the cfHIIT version as realistic as possible. Thus, the racing sections of the ExerCube are equivalent to with the intervals of the cfHIIT and the lower-intensity pitstops are the equivalent for the resting phases of the cfHIIT. However, there are differences in movement sequence (e.g., repetitive movements per block in the cfHIIT vs. varied movements in the ExerCube) and compositions (e.g., different exercises per block in the cfHIIT vs. accumulated exercises over time in the ExerCube"). This should be considered in future work with the ExerCube.

This also includes not artificially removing potentially beneficial social factors from the cfHIIT condition by exploring this in individual sessions instead of groups. However, as the player experience factors were largely higher for the ExerCube condition, a lack of social factors in exergames may not be as much of an issue as expected. However, it should also be noted that participants experienced the ExerCube for the first time in this study, whereas as some of them had prior experience with cfHIIT group classes. The questionnaire results could thus have been influenced by a novelty effect. Future work has to explore whether this effect remains long-term when players become used to the ExerCube, or when the cfHIIT condition is conducted with pre-existing social groups with prior social bonds (our participants did not know each other or exercise together prior to the study).

Another difference lay in the audio-visual scenarios of the stimuli. The ExerCube provides music that adapts to in-game events, with the addition of sound effects for feedback [shown to be important for exergames in previous work (Martin-Niedecken et al., 2019)]. Our cfHIIT stimuli did have comparable music, however, it was not adaptive beyond following and matching the exercise and rest periods. There are some indications that adaptive music can benefit game experiences (Wharton and Collins, 2011; Rogers and Weber, 2019)—and music certainly positively influences exercise (Karageorghis and Priest, 2012)—nevertheless, this has largely not been explored in the exergame context.

When compared to the exergame condition, another difference is that the cfHIIT featured a degree of subjective self-regulation (participants deciding which exercise version they picked, and how fast and intensely to perform them—albeit with guidance from the coach), while the ExerCube featured more objective adjustments (automatically based on the algorithm). We emphasize that we did offer all participants a mobile device positioned on the floor nearby which showed their current HR in real time. As such, they (including the coach) were in theory able to keep track of their individual training, although we cannot report to what degree they used this option.

Finally, the maximal CHR_max in the ExerCube was calculated by a formula that determines relative HR values; these kinds of formulas are based on data from the general population. However, this study was conducted with young healthy adults, whose individual HR_max could potentially be higher than what is predicted by the formula. In future work, we will explore whether the determination of individual HR_max can provide a more customized, higher training intensity without neglecting safety concerns.