Metabolic cost and mechanical work of walking in a virtual reality emulator

Promoting physical activity is critical for preventing and managing chronic diseases and is a focus of major global health initiatives (World Health Organization 2018, 2020; Bull et al. 2020). Virtual reality emulators offer a novel approach to this effort, as they can be combined with a multidirectional treadmill to allow people to walk in a virtual environment while playing active video games or receiving live feedback (Gao et al. 2015; Dębska et al. 2019; Polechoński et al. 2020). In these devices, users glide with their feet on a smooth surface while a harness keeps their body in place, allowing them to freely change their direction and speed. As the users can move freely in a safe manner and receive visual and auditory feedback, multidirectional treadmills have also been suggested as a rehabilitation tool for people with gait disorders (Soni and Lamontagne 2020; Janeh and Steinicke 2021).

Little is known about the energy demands of walking on multidirectional treadmills, but several factors suggest that they may differ from those of normal walking. The harness and platform restrict the users’ movement, which may impair their ability to move their body center of mass (BCoM) in a pendulum-like manner and increase the mechanical work required to raise and accelerate the BCoM (Cavagna et al. 1963, 1976, 1977). However, they may also provide support for body weight during walking, reducing the work done against gravity and hence the metabolic demands (Pavei et al. 2015). Data from Soni and colleagues also show that people walk with a higher stride frequency, shorter swing time, and longer stance time on multidirectional treadmills compared to normal walking (Soni and Lamontagne 2020), which can increase the work done to swing the limbs relative to the BCoM (Fenn 1930; Cavagna et al. 1977; Peyré-Tartaruga et al. 2021). In addition, people adopt a more ‘crouched’ posture on multidirectional treadmills, with higher knee and hip flexion and ankle dorsiflexion during the stance phase (Jochymczyk-Woźniak et al. 2019; Soni and Lamontagne 2020), and lower limb muscle activity (Soni and Lamontagne 2020), which may further increase the cost of walking (Pincheira et al. 2017). Finally, sliding the feet on the treadmill may require additional work to overcome friction along the walking direction.

Determining the energy requirements of walking on multidirectional treadmills would help tailoring their use in preventive medicine and rehabilitation. The World Health Organization recommends that children and adolescents should engage in an average of at least 60 daily minutes of moderate-to vigorous-intensity physical activity—i.e., one which increases the resting metabolic rate by at least threefold—and adults should engage in at least 150 min of moderate or 75 min of vigorous physical activity per week (World Health Organization 2020; Bull et al. 2020). While normal walking is generally a light-to-moderate physical activity, walking on a multidirectional treadmill may be more intense and help meet these recommendations. However, if the metabolic cost of walking on multidirectional treadmills is too high, it may limit the ability to maintain a given walking speed for an extended period of time (Wilkie 1980; di Prampero 1986; Morton 2006), which may limit its use in gait rehabilitation for individuals with reduced aerobic fitness due to advanced age (Fitzgerald et al. 1997; Schneider 2013), disease (Sietsema et al. 2020), or immobilization (Ried-Larsen et al. 2017), while making it suitable for high-intensity interval training (MacInnis and Gibala 2017). Therefore, the aim of the present study was to evaluate the metabolic demands and mechanical work of walking on a multidirectional treadmill used for locomotion in virtual reality.

The metabolic cost of Omni walking was higher than that of normal walking and running—the two most common human gaits—as well as higher than that of skipping, a seldom adopted, highly demanding asymmetric gait. The relationship between C and speed for Omni walking was also U-shaped; however, the minimum of C occurred at a slower speed than for normal walking (Fig. 2). These increased energy demands have implications for the use of multidirectional treadmills in physical activity promotion and rehabilitation. For example, 1 h of active video gaming at a comfortable speed of 0.5 m s⁻¹ would result in a 70 kg player burning 300 kcal above resting values; the same number of calories would be burned when walking for 90 min at 1.4 m s⁻¹ or running for 27 min at 2.8 m s⁻¹. The mean MET was 4.6 ± 1.4, which classifies Omni walking as a moderate-intensity physical activity according to the World Health Organization guidelines (World Health Organization 2020; Bull et al. 2020). Compared to normal walking, this higher intensity could limit the use of walking on multidirectional treadmills for gait rehabilitation in patients for whom such exercise intensity may be too high. Indeed, the metabolic power for Omni walking ranged from 10 to over 30 mlO₂ kg⁻¹ min⁻¹, which may be a high fraction of the maximal aerobic power for untrained individuals and those with conditions impairing their exercise capacity (Fitzgerald et al. 1997; Schneider 2013; Ried-Larsen et al. 2017; Sietsema et al. 2020): this may shorten the amount of time during which such people could walk on multidirectional treadmills. For a comparison, normal walking at self-selected speed requires about 10 mLO₂ min⁻¹ kg⁻¹ (3 METs), with a minimum of 5 mLO₂ min⁻¹ kg⁻¹ (1.4 METs) at 1.3 m s⁻¹ and downhill slope of −18% (Ardigò et al. 2003). On the other hand, the higher metabolic power of walking on multidirectional treadmills could make such activity beneficial to improve cardiorespiratory fitness and the effects of an inactive and sedentary lifestyle when included in physical activity promotion programs and high-intensity interval training interventions (World Health Organization 2020; MacInnis and Gibala 2017; Ekelund et al. 2016). Of note, even though Omni walking has a higher metabolic cost than running and skipping, this does not mean that exercise intensity is higher as well: as shown by the iso-power curves in Fig. 2, running and skipping generally have a higher MET since they are performed at a higher progression speed. The metabolic power required for Omni walking was also strongly correlated with heart rate, with a similar bpm: \( \dot{V}{\text{O}}_{{2}} \) relation to that of normal walking and large between-participant variability in the regression intercepts. These results suggest that heart rate can be used to monitor exercise intensity during walking on multidirectional treadmills (e.g., as done by Dębska et al. 2019), provided that individual differences in resting heart rate are taken into account.

Differences in metabolic demands across and within gaits can often be attributed to differences in mechanical work, since this is one of the main determinants of the metabolic cost (for a review, see Peyré-Tartaruga et al. 2021). In this study we used the same approach by calculating the ‘traditional’ sources of mechanical work when walking on the multidirectional treadmill; however, differences in W_TOT alone did not fully explain the differences in C. Indeed, W_TOT decreased linearly with speed, whereas C followed a curvilinear trend. Moreover, the efficiency of Omni walking was consistently lower than that of normal walking: it would have been the same if W_TOT had explained all the increase in C. Analyzing the components of W_TOT can shed further light on the mechanisms underlying its variations between gaits. External mechanical work decreased with speed for Omni walking while it increased for normal walking, with an increasing energy recovery for both gaits. As W_EXT can also be considered as (1−Recovery)(W_forward + W_vertical) (Luciano et al. 2022), where (W_forward + W_vertical) is the sum of the increases in the kinetic energy of the forward motion and the vertical energy of BCoM, respectively, such opposite speed-dependency for W_EXT is due to a divergent trend for this last term, rather than for the ability to behave in a pendulum-like manner. On the other hand, differences in W_INT,k can be explained by differences in limb kinematics. At a given speed, W_INT,k is proportional to f(1 + (d/(1−d))²) (Minetti and Saibene 1992; Minetti 1998), where f is the stride frequency and d is the duty factor: at lower speeds, stride frequency and duty factor were higher for Omni walking than normal walking, but at higher speeds the differences between the two activities tended to disappear (Fig. 6). As a result, W_INT,k was higher in multidirectional treadmill walking only at speeds below 1 m s⁻¹. Of note, in this study we compare Omni walking at imposed frequencies with reference data on treadmill walking at fixed speeds. However, observed differences in cost, mechanical work, and kinematics are largely greater than those attributable to inter-participant variability and variations in stride frequencies (Minetti et al. 1995; Umberger and Martin 2007; Stoquart et al. 2008; Pavei et al. 2015).

While the work against external frictions is negligible during normal walking, it may be a major determinant of the high metabolic cost of walking on a multidirectional treadmill. Indeed, apparent differences in efficiency between the two gaits were largely reduced when the work done against sliding friction was taken into account. Some further factors were not included in the present analysis, but may contribute to the high cost of walking on the multidirectional treadmill: among them, internal frictional mechanical work (Minetti et al. 2020) done within joints. Such term was not included in W_TOT since its partitioning with W_INT,k has not been solved yet, and little can be inferred about differences in internal frictional mechanical work between Omni and normal walking. Additionally, the denominator of efficiency is given by metabolic cost, which also includes the energy spent by muscles that do not perform mechanical work, such as those contracting isometrically or co-contracting. When walking on a multidirectional treadmill, a greater fraction of active muscles may be activated to comply with stability and postural needs due to the more crouched posture—as hinted also by the higher electromyographic activity of the lower limbs reported by Soni and Lamontagne (2020)—without generating mechanical work. Finally, during normal walking, some energy is saved due to storage and release of elastic energy by the tendons and connective structures of the hip and ankle joints (Fukunaga et al. 2001; Eng et al. 2015) and due to optimized muscle and tendon gearing of knee extensors and plantar flexors (Monte et al. 2022); however, the different kinematics of such joints during Omni walking may hamper these energy-saving mechanisms.

The unique kinematic patterns, BCoM motion, and limb mechanics of walking on multidirectional treadmills distinguish it from normal walking and highlight the need for caution when evaluating its use in gait rehabilitation. Compared with normal walking, Omni walking showed a more flexed trunk, a dorsally flexed ankle, and a different angular displacement at the hip and ankle joints. Moreover, participants attained slower speeds than in normal walking and with different spatiotemporal parameters, as previously observed by other authors (Jochymczyk-Woźniak et al. 2019; Soni and Lamontagne 2020). Together with differences in mechanical work and metabolic cost, such observations challenge the ability of multidirectional treadmills and virtual reality emulators to mimic normal walking. Further studies should test whether observed features of walking on the Omni treadmill can be generalized to other multidirectional treadmills. We can speculate, however, that this should be the case for at least some of them. For instance, also Soni and Lamontagne (2020) found that, compared to normal walking, people had a lower stride length when walking on a different multidirectional treadmill. Moreover, all passive treadmills require to perform work against sliding frictions, and the coefficients of these frictions should be maintained within a narrow range to prevent surfaces from becoming unrealistically slippery.

Walking on a multidirectional treadmill has a higher metabolic cost and different mechanics and kinematics than normal walking. These findings question the use of multidirectional treadmills in rehabilitation when the goal is to mimic normal gait, but support their use to promote high-intensity exercise and physical activity in the general population.