Overground to omnidirectional: Spatiotemporal and muscle activation pattern adaptations across walking modalities,virtual reality,and pace constraints

Abstract

The rapid expansion of virtual reality (VR) technology has led to the development of low-friction, slip-style omnidirectional treadmills (OTs), which have great promise for implementation into VR-based gait rehabilitation protocols. However, previous work indicates that use of these treadmills leads to unique gait patterns that may differ from overground gait, and there is a lack of research examining how OT gait differs from a conventional, belt-driven treadmill (CT). Thus, the purpose of this study was to characterize spatiotemporal and electromyographic gait patterns on an OT and to compare them with overground and CT walking. Fourteen healthy participants walked in these three conditions in both the real-world and in VR, as well as at fixed and preferred paces. Results indicate that use of the CT promoted significantly longer stride lengths (mean = 1.838 m) and reduced variability (coefficient of variation (CV) = 14.7%) compared to overground walking (mean = 1.578 m, CV = 21.4%). Moreover, the use of CT in VR at a fixed pace led to reduced biceps femoris (CV = 20.4%) and medial gastrocnemius (CV = 14.9%) variability compared to overground walking (biceps femoris CV = 26.4%, medial gastrocnemius CV = 22.2%), while use of the OT demonstrated variability similar to that of overground walking across measures. These results indicate that the user-driven aspect of OTs may elicit gait patterns more similar to overground walking than traditional belt-driven treadmills.

Keywords

virtual reality omnidirectional treadmill electromyography inertial measurement units gait variability cadence stride length biceps femoris gastrocnemius pacing

Background

Virtual reality (VR) is a rapidly-growing tool for physical rehabilitation¹ that allows for patients to practice in an environment customized to their needs (e.g., difficulty)^2,3 without the physical space and equipment demands that conventional rehabilitation may require.^3,4 Recent work indicates that VR provides other benefits, such as a high level of motivation and engagement relative to conventional rehabilitation.^3,5,6 Moreover, the current generation of VR systems introduce immersive, high-resolution environments that provide realistic training scenarios to positively facilitate real-world skill transfer.^6–8 More recent generations of VR systems have become more user-friendly; many newer systems are wireless (i.e., no physical tethering to a PC)⁹ and use hand recognition in place of controllers (allowing for more realistic performance of manual tasks).^6,10 Accordingly, many recent VR interventions have noted positive rehabilitation outcomes across a variety of skills and pathologies, including upper-extremity training for those with stroke,¹¹ balance training in patients with traumatic brain injury,¹² and functional improvement in individuals recovering from orthopedic surgery.¹³

While the aforementioned findings show promise for further implementation of VR into rehabilitation clinics, gait rehabilitation in VR faces several challenges that require further scientific inquiry before these benefits can be maximized. For one, cybersickness from VR use occurs frequently in locomotor tasks, often characterized by nausea, disorientation, and eye strain (among other symptoms).¹⁴ One common explanation for the prevalence of cybersickness while walking in VR is the difference in sensory inputs between the real and virtual worlds; if visual and vestibular information experienced in VR do not align, then cybersickness can occur.^15,16 Second, the level of immersion, or the amount of realism present in a virtual environment, can influence transfer of a learned locomotor skill from VR to the real world.^17,18 Overall, researchers recommend that a virtual environment designed for locomotor training should replicate real-world gait as closely as possible while mitigating sensory conflicts and maximizing immersion.¹⁹ However, overground walking in VR requires the same amount of space as the real world, posing the unique challenge of what McGrath Lewis et al.²⁰ have termed the ‘VR locomotion problem’, where navigating a large virtual environment (e.g., a grocery store) within a smaller real-world space (e.g., a clinic or research laboratory) overground is not possible.

To address the VR locomotion problem, researchers and clinicians have integrated conventional (unidirectional) treadmills (CTs) into their VR systems. While CTs address the VR locomotion problem in terms of space requirements, users are relegated to a single direction of motion, thereby limiting skill transfer to the more complex, multi-directional real-world walking.^20,21 More recently, omnidirectional treadmills (OTs) have been introduced into commercial markets to allow for unconstrained movement in three-dimensions in VR.²² OTs are typically used for gaming applications,²³ and are either belt-driven (e.g., Infinadek (Rocklin, CA USA) or low-friction, slip-style (e.g., Virtuix Omni One (Austin, TX USA)). The slip-style treadmills typically require the user to don low-friction shoes and walk by sliding their feet on a concave surface. While both types of OTs are user-paced, low-friction systems are much less complex and expensive and are therefore more accessible to the lay population.²⁴

Nevertheless, recent work examining gait kinematics on low-friction OTs has demonstrated that OT walking is different from overground gait, with users exhibiting faster cadence,²⁵ shorter step lengths,²⁰ decreased speed,²⁰ longer stance phases,²⁶ increased stepping variability,²⁰ and higher metabolic cost²⁷ during straight-line walking. These differences in gait kinematics are logical when the design of the slip-style OT is considered, as the concave surfaces would decrease a user’s stride length if they first struck the raised edge during initial contact. Moreover, this treadmill design requires the user to lean forward into the harness in order to slide their feet behind them²⁶ (similar to that of an ice-skating motion). Thus, the forward body lean of OT users is expected to differ from that of the typical upright posture implemented by healthy overground walkers. As a result, leg muscle recruitment patterns deviate from typical overground gait to accommodate this posture. For example, Soni and Lamontagne described increased hip extensor and hip flexor amplitude in walkers on OTs compared to overground.²⁵ They attributed this adaptation to either co-contraction activity occurring as a result of the increased challenge of OT walking or new muscle synergies.²⁵ Thus, the emergence of novel muscle recruitment patterns in combination with differing kinematic strategies, demonstrate the substantial differences between OT walking and overground gait. This finding brings into question the utility of OTs as a rehabilitative tool, as well as the ability of OT users to transfer learned gait patterns to real-world walking.

Multiple studies compare OT walking to overground walking^20,25 and Bashir et al.²⁴ have compared OT walking with CT and overground walking on user experience and cybersickness outcomes. However, there is still a need to characterize kinematic and electromyographic (EMG) outcomes during OT walking in relation to CT walking. As many clinics employ CTs for gait rehabilitation, it is important to characterize the difference in walking patterns between the two types of treadmills to determine if the added benefit of free, self-paced movement on an OT outweighs the more familiar form of walking on a CT.²⁸ It is also unknown whether the previously noted kinematic and EMG differences between treadmill types reflect the physical constraints of the treadmill or the influence of VR. Some researchers have noted that being immersed in VR while walking may elicit a more conservative gait,^29,30 particularly in older populations.³¹ In addition, researchers have noted that many OT users experience cybersickness, which naturally would dampen adoption of OT in future clinical applications.³² The interaction between cybersickness and OT use may be related to pace, where faster walking paces may induce greater visual-vestibular sensory conflicts.³² Thus, differences in kinematic and EMG patterns implemented during OT walking compared to CT and overground walking may be related to whether VR is being used and walking pace.

Therefore, the purpose of this study was to compare gait kinematics, variability, and muscle activation patterns in healthy participants walking on OTs, CTs, and overground. Participants completed these walking conditions while being both in VR and in the real world, and also at both their preferred pace and a fixed pace calculated based on their preferred overground gait.

We hypothesized that:

• Hypothesis 1a: walking on the OT would result in shorter stride lengths, faster cadences, and increased variability in these outcomes compared to CT and overground conditions.

• Hypothesis 1b: walking at a fixed pace would result in decreases to stride length across conditions.

• Hypothesis 1c: walking in VR would produce a more uncertain gait, characterized by a decreased stride length, decreased cadence, and increased variability across measures.

• Hypothesis 2a: while using the OT, greater hip extensor (biceps femoris; BF) activity and lesser plantar flexor (medial gastrocnemius; MG) activity, as well as increased variability in these activation patterns would be observed.

• Hypothesis 2b: this change in EMG activity and variability would be magnified (i.e., even greater differences) at a fixed pace.

• Hypothesis 2c: this change in EMG activity and variability would also be magnified by VR use.

Methods

Participants

Fourteen young, healthy individuals (8 males, 6 females, 25.8 ± 3.7 years old) participated in this study. This sample size was estimated using G*Power 3.1.9.7³³ with a large effect size noted by Bashir et al.²⁴ on step count differences between OT and CT treadmills. This analysis indicated that a minimum of 12 participants was needed to achieve power (β) = 0.95 and α = 0.05. This sample size is also aligned with similar studies that examined gait outcomes during OT use.^20,25 Eligibility to participate included being between the ages of 18 to 65, having no acute injury within the last three months, and having no chronic musculoskeletal injuries, cardiovascular, and/or neurological disorders that affected gait. Individuals were excluded if they had abnormal vision that could not be corrected with contact lenses, impaired hearing without the use of hearing aids, or if they were unable to walk for 45 consecutive minutes. Participants who wore glasses to correct their vision were excluded from participation as wearing glasses underneath a VR headset can cause discomfort and distract participants from the gait task. All participants provided written informed consent prior to participation. The protocol adhered to the Declaration of Helsinki and was approved by the university’s Institutional Review Board (IRB).

Experimental procedure

All participants completed 12 walking conditions. These 12 conditions included three different treadmills/surfaces: 1) overground (Figure 1(A)), 2) a unidirectional, belt-driven CT (Precor C954, Woodinville, WA USA; Figure 1(B)), and 3) a slip-style omnidirectional treadmill (Virtuix Omni, Austin, TX USA; Figure 1(C)). Within each treadmill condition, participants walked at both 1) preferred and 2) fixed cadences, and in both 1) real-world and 2) VR environments. Each condition was repeated three times for seven m (overground) or 30 seconds each (treadmill conditions), totaling 36 trials per participant (Table 1).

Figure 1.

Overview of walking conditions (A) overground, (B) on a conventional treadmill, and (C) on an omnidirectional treadmill. (D) View of OctonicVR app from user perspective.

Table 1.

Overview of walking conditions and associated abbreviations.

Treadmill/Surface	Walking cadence	Environment	Abbreviation
Overground	Preferred	No VR	OPN
	Preferred	VR	OPV
	Fixed	No VR	OFN
	Fixed	VR	OFV
Conventional	Preferred	No VR	CPN
	Preferred	VR	CPV
	Fixed	No VR	CFN
	Fixed	VR	CFV
Omnidirectional	Preferred	No VR	VPN
	Preferred	VR	VPV
	Fixed	No VR	VFN
	Fixed	VR	VFV

The OPN condition was always performed first to capture participants’ natural speed and cadence as the “baseline” condition. During the OPN condition, the time it took to complete the 10 m walk was recorded as well as the number of steps taken for each trial. Then, each participant’s preferred walking speed (m/s) was calculated as the distance (10 m) divided by the average trial time across the three trials (s). Cadence (steps/min) was estimated as the number of steps taken divided by the average trial time (min). This information was used for setting fixed paces like the treadmill speed in the CT condition and metronome in the overground and OT conditions (see below). Next, all remaining overground conditions (OPV, OFN, OFV) were completed in a random order with three trials for each condition performed consecutively before proceeding to the next condition.

Following completion of the overground conditions, the first treadmill type (CT or OT) was selected at random, with participants completing all pace and VR conditions for that treadmill in a randomized order before participants moved to the other treadmill. This randomization procedure was selected to minimize the number of transitions between treadmills while still minimizing any fatigue or learning effects that may have occurred during later conditions. In the CT and OT conditions, participants were given five seconds to acclimate to the treadmill before data were recorded. To find participants’ preferred speeds on the CT (CPN, CPV), a member of the research team gradually increased the treadmill speed until the participant notified them that they were at a comfortable speed. Fixed speeds on the CT (CFN, CFV) were set to the rounded calculated speed recorded during the OPN condition, and participants were instructed to synchronize their steps to each beat of a metronome, which was determined from their calculated cadence in the OPN condition. As the OT was self-paced, the fixed conditions (VFN, VFV) were set to participants’ calculated cadence rather than speed with a metronome (in BPM). Participants were instructed to synchronize their steps to each beat of the metronome. For preferred speeds (VPN, VPV), participants were asked to walk at a comfortable pace.

Participants wore their own shoes for all conditions; however, during the OT condition, they also wore specialized low-friction overshoes that fit over their personal footwear. For the OT condition, they were fitted into the treadmill’s ring harness system that was vertically adjusted to the height of their pelvis. Participants were given a familiarization period to practice walking, as all participants had no prior experience using an OT.

For all VR trials, each participant was fitted with a Meta Quest 2 (Menlo Park, CA USA) head-mounted VR system that ran Octonic VR (New York, NY USA), a commercially-available application developed to allow for free movement and exercise in various virtual environments. The virtual environment used for this study consisted of an oval track in a stadium, where participants could see a small avatar in front of them that mirrored their stepping rate in real time (Figure 1(D)). This application was chosen as it has both a non-treadmill mode, allowing for free movements in the real world (used for overground conditions), and a treadmill mode, enabling movements in the virtual environment without forward motion in the real world. In non-VR trials, participants removed the headset and were not given specific instructions regarding where to direct their gaze.

Data collection

Prior to beginning the walking trials, participants were instrumented with four force-sensitive resistors (FSRs; Noraxon, Scottsdale, AZ USA) on the bottom of each foot over participants’ socks (Calcaneus, 1^st metatarsal, 5^th metatarsal, hallux). A second pair of socks were placed over the FSRs to secure them. These resistors were used to detect gait events such as initial contact and toe off, which allowed for separation of strides when calculating outcome variables (see Data Analysis). In addition, an inertial measurement unit (IMU; Noraxon, Scottsdale, AZ USA) was placed on the dorsal aspect of each foot, allowing for calculation of kinematic variables. These data were recorded at 300 Hz.

Additionally, participants were equipped with two bipolar surface EMG electrodes (Noraxon, Scottsdale, AZ USA) on their dominant leg (i.e., the leg they would use to kick a ball): one over the biceps femoris (BF) and one over the medial gastrocnemius (MG). The BF was chosen as it was subjectively noted during preliminary testing that more forceful hip extension was required in order to slide the feet backward when using the OT. The hip extensors also strongly contribute to forward propulsion during the stance phase of gait to compensate for decreases in force generation from the ankle plantar flexors.³⁴ As the frictionless surface of the OT may limit plantar flexor power generation, compensatory adaptations by the hip extensors would be apparent. The MG was chosen because prior work in gait adaptations have studied this muscle as a proxy measure for force generation during the toe-off phase of gait.³⁵ This muscle has also been recently studied in other work examining gait on OTs.²⁵ EMG data were recorded at 1500 Hz.

Data analysis

Anterior/posterior linear accelerometer signals recorded from IMUs were low-pass filtered using a zero-lag, 4^th-order Butterworth filter with a 75 Hz cut-off frequency, and then high-pass filtered at 0.5 Hz using a zero-lag, 4^th-order Butterworth. These signals were then used to calculate the following outcome variables. Stride length was computed as the distance the foot accelerometer traveled from toe off on one foot to the subsequent initial contact on the same foot. Distance was calculated through double numerical integration of the anterior/posterior accelerometer signal. To account for drift, linear acceleration signals were separated into individual swing phases from the whole time series using FSR data (toe off to subsequent heel strike), zeroing the initial velocity for overground and OT trials prior to numerical integration. This method, combined with the integrated sensor fusion algorithms implemented in the Noraxon MyoMotion (Noraxon, Scottsdale, AZ USA) software that mitigate gravity effects, effectively provided a ‘hard’ zero-velocity update for each stride, which has been previously described in detail by Skog et al.³⁶ For CT trials, the anterior-posterior velocity of the foot at the instant of toe off was used as the initial velocity during this calculation to account for the treadmill belt speed. Cadence was estimated as the number of steps taken in one minute, as measured through registered heel strikes by the force sensitive resistors on the heels. Stride length and cadence were both averaged across steps for each trial and condition, and their coefficients of variation (CVs) were calculated as the standard deviation for each trial divided by the mean.³⁷

EMG signals were band-pass filtered from 50-450 Hz, then the linear envelope was obtained by low-pass filtering these data at 5 Hz using a 4^th-order Butterworth filter. The linear envelope was then used to find the integral of the EMG signal during stance phase (identified using the force sensitive resistors on the heel and toe) for each step during each trial. EMG activity was normalized to the integrated signal from the overground, preferred, no VR condition (baseline). As such, all EMG values were expressed as a percentage of activity during a typical gait with no VR. Then, these normalized values were averaged across trials and conditions, and their CVs were also calculated in the same manner as stride length and cadence. All data analysis was performed using MATLAB 2023b (Mathworks, Natick, MA USA).

Statistical analysis

To test for the main effects of treadmill type (overground, CT, OT), VR use (VR or no VR), and pace (preferred or fixed), as well as interactions between these main effects, on stride length, cadence, BF EMG activity, and MG EMG activity, a repeated-measures multivariate analysis of variance (rMANOVA, α = 0.05) was performed. A second rMANOVA (α = 0.05) was performed to test for the same main effects and interactions on each dependent variable’s respective CV. For both analyses, follow-up Sidak-corrected pairwise comparisons were made if a main effect or interaction was revealed to be statistically significant. Prior to these analyses, assumptions of normality and sphericity were checked. If the assumption of sphericity was violated for a given variable, Greenhouse-Geisser corrections were used. All statistical analyses were done using SPSS v28 (IBM, Armonk, NY USA).

Results

There was a significant main effect of treadmill type on stride length, with a large effect size (F = 8.791, p = 0.027, $η_{p}^{2}$ = 0.403; Figure 2(A)–(D)). Follow-up pairwise comparisons indicate that walking on the CT elicited a significantly longer stride length than both overground (p = 0.008; 95% CI [0.067 0.452]) and OT (p = 0.010, 95% CI [0.153 1.127]; Figure 2(A)–(D)). There were no significant main effects of pace (F = 0.669, p = 0.628, $η_{p}^{2}$ = 0.211; Figure 2(A)–(D)) nor VR (F = 1.688, p = 0.229, $η_{p}^{2}$ = 0.403; Figure 2(A)–(D)), nor were there any significant interaction effects.

Figure 2.

Stride length (m) and stride length coefficient of variation (%) while walking overground, on a conventional treadmill, and on an omnidirectional treadmill. (A) & (E) Walking at a preferred pace without using VR. (B) & (F) Walking at a preferred pace while using VR. (C) & (G) Walking at a fixed pace without VR. (D) & (H) Walking at a fixed pace while using VR.

There was also a significant main effect of treadmill on stride length CV, with a large effect size (F = 3.681, p = 0.045, $η_{p}^{2}$ = 0.221; Figure 2(E)–(H)). Follow-up comparisons indicate that CT walking resulted in a significantly lower stride length CV while not in VR at a fixed pace compared to overground walking (p = 0.012, 95% CI [1.982 16.717]; Figure 2(E)–(H)).

There were no significant main effects of treadmill on cadence, cadence CV (Figure 3), BF EMG (Figure 4(A)–(D)), nor MG EMG (Figure 5(A)–(D)).

Figure 3.

Cadence (steps/min) and cadence coefficient of variation (%) while walking overground, on a conventional treadmill, and on an omnidirectional treadmill. (A) & (E) Walking at a preferred pace without using VR. (B) & (F) Walking at a preferred pace while using VR. (C) & (G) Walking at a fixed pace without VR. (D) and (H) Walking at a fixed pace while using VR.

Figure 4.

Biceps femoris activity (normalized to overground, preferred, no VR) and biceps femoris coefficient of variation (%) while walking overground, on a conventional treadmill, and on an omnidirectional treadmill. (A) & (E) Walking at a preferred pace without using VR. (B) & (F) Walking at a preferred pace while using VR. (C) & (G) Walking at a fixed pace without VR. (D) and (H) Walking at a fixed pace while using VR.

Figure 5.

Medial gastrocnemius activity (normalized to overground, preferred, no VR) and medial gastrocnemius coefficient of variation (%) while walking overground, on a conventional treadmill, and on an omnidirectional treadmill. (A) & (E) Walking at a preferred pace without using VR. (B) & (F) Walking at a preferred pace while using VR. (C) & (G) Walking at a fixed pace without VR. (D) and (H) Walking at a fixed pace while using VR.

There was, however, a significant main effect of treadmill on MG CV, with a large effect size (F = 9.539, p = 0.004, $η_{p}^{2}$ = 0.423; Figure 5(E)–(H)). Pairwise comparisons found a significantly greater MG CV when walking on the OT compared to CT (p = 0.006, 95% CI [5.010 28.723]; Figure 5(E)–(H)). Furthermore, there was a significant main effect of pace on both BF CV (F = 11.134, p = .005, $η_{p}^{2}$ = 0.461; Figure 4(E)–(H)) and MG CV (F = 19.343, p = 0.001, $η_{p}^{2}$ = 0.598; Figure 5(E)–(H)). Follow-up comparisons indicated significantly greater variability at fixed paces compared to preferred paces for both BF CV (p = 0.005, 95% CI [2.518 11.765]; Figure 4(E)–(H)) and MG CV (p = 0.001, 95% CI [6.470 18.961]; Figure 5(E)–(H)). Additionally, there was a significant multivariate effect of VR use on the CV variables with a very large effect size (F = 4.073, p = .033, $η_{p}^{2}$ = 0.620), yet no significant pairwise comparisons were revealed. Thus, this analysis was performed a second time with Fisher’s LSD. This revealed that VR induced greater BF CV (p = 0.005, 95% CI [2.518 11.765]; Figure 4(E)–(H)) and MG CV (p < 0.001, 95% CI [6.470 18.961]; Figure 5(E)–(H)) compared to no VR.

Moreover, there was a significant treadmill*pace*VR interaction with a very large effect size (F = 4.324, p = 0.046, $η_{p}^{2}$ = 0.852). Follow-up pairwise comparisons indicate that for BF and MG CV, CT (p < 0.001, 95% CI [7.780 30.397]) and OT (p < 0.001, 95% CI [10.320 26.160]) had greater variability than overground when walking at a preferred pace without using VR (Figures 4 and 5). When VR was used, walking on a CT produced an opposing effect, where BF CV was lower on the CT compared to overground (p < 0.001, 95% CI [7.904 21.757]; Figure 4(F)). When walking overground (p < 0.001, 95% CI [20.015 35.461]) and on the OT (p = 0.050, 95% CI [-0.007 17.891]) when not using VR, walking at a fixed pace resulted in a significantly greater BF CV compared to that of a preferred pace (Figure 4(G)). Finally, when walking overground at a preferred pace the use of VR significantly increased BF CV compared to no VR (p < 0.001, 95% CI [19.519 33.497]; Figure 4(E) and (F)).

Similar to BF CV, in the preferred no VR condition walking on a CT (p < 0.001, 95% CI [7.609 16.289]) and OT (p = 0.014, 95% CI [4.417 41.560]) resulted in a significantly higher MG CV compared to overground walking (Figure 5(E)). In addition, while in VR at a fixed pace, walking on the OT resulted in a significantly higher MG CV compared to CT (p < 0.021, 95% CI [4.153 53.719]; Figure 5(H)). For all treadmill conditions while using VR, walking at a fixed pace resulted in a significantly greater MG CV than a preferred pace (overground: p < 0.001, 95% CI [17.813 49.110]; CT: p = 0.001, 95% CI [2.995 9.913]; OT: p = 0.045, 95% CI [0.436 30.757]; Figure 5(E)–(H)). On the other hand, while not in VR, there was no significant difference in MG CV across treadmill conditions while walking at a fixed pace. Finally, in the overground condition walking in VR at a preferred pace produced a significantly greater MG CV than no VR (p < 0.001, 95% CI [15.850 29.118]; Figure 5(E)–(H)).

Discussion

The purpose of this study was to characterize spatiotemporal and electromyographic differences in gait among healthy adults walking on an omnidirectional treadmill (OT), a conventional treadmill (CT), and overground. It was hypothesized that (Hypothesis 1a) walking on the OT would result in significantly shorter stride lengths, a faster cadence, and increased muscle activity of the biceps femoris and medial gastrocnemius compared to the other two treadmill conditions. Additionally, we hypothesized that (Hypothesis 1b) walking at a fixed pace would decrease stride length, and that (Hypothesis 1c) walking in VR would decrease stride length, decrease cadence, and increase BF and MG activity. It was also hypothesized that (Hypothesis 2a) walking on the OT would result in greater variability in these outcome variables, as measured through the coefficients of variation. Moreover, we expected even greater variability during OT walking in conditions where participants walked at a fixed pace (Hypothesis 2b) and in VR (Hypothesis 2c).

In general, the results of this study do not support Hypothesis 1a, as the OT condition did not produce any significantly different stride length, pace, or BF/MG activity outcomes compared to overground. Instead, walking on the CT resulted in significantly greater stride lengths compared to the other two conditions, and no significant differences existed between OT and overground walking. This result contrasts with some recent work examining the effect of OT walking on step/stride length. For example, Soni and Lamontagne²⁵ noted a shorter step length during OT walking compared to overground, which they attributed to a compensation occurring to ensure balance on the OT. They speculated that the low-friction surface may have reduced stability and their ability to push off during late stance.²⁵ Souman et al.³⁸ also found that stride lengths were shorter on an OT, albeit with a belt-driven treadmill as opposed to a slip-style system as used by both Soni and Lamontagne²⁵ and the present study. On the other hand, McGrath Lewis et al.²⁰ found no differences between OT and overground walking on step length while using a belt-driven treadmill. Thus, the present study aligns with that of McGrath Lewis et al.,²⁰ and to date is the only study that demonstrates that slip-style OTs do not significantly alter stride length from that of overground walking. Still, more research across different populations and treadmill types is needed to determine why there are inconsistent results across studies in terms of changes to step/stride length during OT walking.

The results of this study also do not support Hypothesis 1b, as there were no significant effects of pace (preferred/fixed) on spatiotemporal or EMG outcome variables. The similarity in outcomes between fixed and preferred pace walking presents a direct contrast to the work by Soni and Lamontagne, who found that on a low-friction OT changes in walking speed led to changes in both step length and cadence.²⁵ Outside of OT walking, other more general human gait research has described how the relationship between stride length and cadence are tightly regulated by automatic neural processes during walking at a self-selected speed (e.g., the preferred pace condition in the present study).³⁹ However, more voluntary control mechanisms exist when walking in a novel context (such as on an OT), eliciting an inconsistent relationship between stride length and cadence.³⁹ Therefore, in the present study, it follows that on an OT stride length would be affected when pace was constrained (i.e., fixed pace). However, the results of this study do not support this theoretical outcome. More research is warranted to uncover whether stride length-cadence coupling is affected differently during OT walking compared to overground.

Similarly, the finding that the use of VR did not affect gait outcomes was surprising in comparison with prior literature, which does not support Hypothesis 1c. As previously discussed, researchers have found that walking in VR produces a more conservative gait (e.g., reduced stride length) compared to walking without VR.²⁹ Other related work has also described how gait kinetics, such as peak force generation during toe-off, is also impacted by VR use.³⁰ Some researchers have attributed participants’ ability to replicate their natural gait in VR to the realism of a virtual environment, particularly with respect to accuracy of self-motion perception in the virtual environment.⁴⁰ In other words, if a VR user does not correctly estimate their walking speed due to the design of a virtual environment (such as incorrect perceptual velocity cues from inaccurate optic flow or incorrect virtual avatar movement) their gait will deviate more widely from real-world walking. Thus, it is possible that in the present study, the virtual environment selected adequately mirrors real-world visual feedback during walking, given that no kinematic changes to gait were observed between VR/no VR conditions. This finding highlights the importance of realistic virtual environment physics when designing gait rehabilitation interventions using an OT, as virtual environments that closely mirror the real world will be less likely to impact gait.

While very few absolute changes to gait kinematics and muscle activity were seen across walking conditions, there were large differences in gait variability outcomes between conditions. For one, there was a main effect of treadmill on stride length CV, and pairwise comparisons indicated a significantly lower amount of variability on the CT compared to overground walking. This finding partially supports Hypothesis 2a and is aligned with previous research.⁴¹ Hollman et al.⁴¹ cautioned that smaller variability may not always be a positive outcome and may be indicative of a lack of adaptability, as seen in older populations and those with movement disorders. Following this logic, the lack of significant spatiotemporal gait variability differences between overground and OT walking demonstrates that OT walking may promote a level of variability comparable to overground gait, highlighting its potential ability to better mimic overground walking compared to the use of a CT. More research examining long-term adaptations to OT walking may further elucidate whether spatiotemporal variability mirrors that of overground gait, or if adaptations over time reduce variability to a level more comparable to that of CT walking. Further, more research comparing belt-driven to slip-style OTs is warranted, as the self-selected gait inherent in the slip-style OT may drive the natural variability present in overground gait.

Greater amounts of variability were present in MG activation patterns on the OT compared to the CT, which partially supports Hypothesis 2a. The plantar flexor muscles provide reflexive, feedback-based adaptations during late stance to provide push-off power.⁴² Ogawa et al. have attributed the magnitude of such adaptations during treadmill walking to the speed of the treadmill belt (rather than a predictive, centrally-mediated feedforward approach).⁴² To speculate, on the slip-style OT, little friction is available to push off the ground, which limits the ability of the MG muscles to generate power. Instead, the hip flexors may take on a greater role to pull the trailing limb into swing phase. Thus, the high amounts of MG variability may be a downstream consequence of exploratory stepping patterns as participants acclimated to the new treadmill. Furthermore, the specific slip-style treadmill used in this study implemented a ring harness around the participant’s pelvis. A natural consequence of this harness design is that participants’ thighs often contacted this ring if their leg moved too vertically during swing phase, which likely drove greater variability in muscle activity. Thus, the short familiarization and walking trials on the OT in the present study were likely insufficient for participants to fully acclimate, laying out another reason for future research to explore longer-term adaptations.

Higher variability in muscle activation patterns was also observed at fixed paces specifically when VR was used. At a fixed pace, both BF and MG activity were more variable across conditions, which supports Hypothesis 2b. This increased variability due to fixed pacing aligns with previous work, which has proposed that fixed pacing imposes mechanical constraints that do not promote a natural gait.^41,43 The interaction between treadmill type, pace, and VR use seems to alter this relationship, as CT and OT walking at a preferred pace had greater variability in muscle activity patterns when VR was used. However, the inverse occurred when VR was implemented, as CT walking had lower variability in muscle activity compared to overground, and the patterns between CT and OT walking were not significantly different. Thus, it is possible the perceptual cues (e.g., optic flow) presented in VR may have not aligned with the real world, causing an uncertain gait during overground walking—a phenomenon which did not occur on the two treadmills where optic flow out of VR is not present.⁴⁰ This finding is supported when comparing VR to no VR conditions during overground walking, where BF and MG variability was greater in the VR condition. Therefore, Hypothesis 2c is supported by the data. In total, VR may have a detrimental effect to overground gait, lending credence to the use of treadmills in clinical gait rehabilitation procedures.

Overall, use of a slip-style OT produces gait patterns similar to that of overground walking than CT walking. To posit, the belt-driven CT may rely on a more feedback-based neural control mechanism, which when at a fixed speed (as in the present study), the gait cycle becomes constrained and limits exploration of new walking strategies.⁴⁴ From a clinical perspective, when using VR in gait rehabilitation, an OT may allow for better transfer of learned gait patterns to real-world, overground walking.

However, more research across multiple domains is needed to further clarify how gait is affected by use of OTs. First, further study is needed to examine the direct transfer from OT walking to overground following repeated practice. In addition, gait kinematics (e.g., leg joint angles) and kinetics (e.g., leg joint moments) were not measured in this study; additional research is needed to confirm if these gait strategies are similar between OT walking and overground before OT training can be recommended. Moreover, there is a need to better understand the user experience during OT walking. Anecdotally, in the present study many participants found use of the OT to be awkward and fatiguing; however, we did not administer a formal survey to capture those perceptions. Thus, it is crucial for future work to gather user perspectives (e.g., questionnaires or semi-structured interviews) regarding OT walking before these treadmills are recommended clinically. As this study was conducted on a sample of healthy individuals, future work should examine how gait is altered in those with movement disorders (e.g., stroke, Parkinson’s disease) when walking on the OT. Finally, those who wore glasses to correct their vision were excluded from this study, which may limit the applicability of these findings to many undergoing rehabilitation, particularly older adults. While some prototype devices exist that provide eye correction through the headset,⁴⁵ these systems are (to date) not commercially available. Therefore, the full adoption of slip-style OTs in rehabilitative settings may be limited until these practical challenges are fully addressed.

Conclusion

VR’s ability to be personalized to user needs, as well as minimizing space and equipment requirements, makes it a very promising application for gait rehabilitation. When VR is implemented on an OT, people walk similarly to that of overground from a spatiotemporal, EMG, and gait variability perspective. Conversely, CT walking produces differing gait patterns from OT and overground walking. This effect is magnified by the use of fixed pacing. Hence, OTs may facilitate improved transfer to overground walking compared to CTs as they produce a gait pattern close to that of overground walking, possibly because these treadmills are user-driven rather than by a belt. Before OTs can be fully recommended for clinical use, more research examining direct transfer to overground walking, leg joint kinematics, and user experiences is needed. Crucially, future research should replicate this work examining how those with movement disorders such as stroke walk on OTs, as adopting a novel gait may be more difficult for these populations. Nevertheless, as scientific understanding of slip-style OTs continues to develop, their applications for rehabilitation have great promise to meet the needs of users and clinicians/practitioners alike.

Footnotes

Acknowledgments

Thanks to Rabbani Nzeza and the other members of the Gait, Oculomotor, and Lower-extremity Dynamics (GOLD) Lab for assistance with project development and data collection.

ORCID iD

Jacob W. Hinkel-Lipsker

Ethical considerations

All study procedures were approved by the Institutional Review Board at California State University, Northridge (#IRB-FY23-176).

Consent to participate

All participants provided written informed consent prior to enrolling in the study.

Author contributions

W.T.P.- Pilot testing, data collection and analysis, manuscript writing and editing.

B.A.B.- Pilot testing, data collection and analysis, manuscript writing and editing.

K.E.H.- Pilot testing, data collection and analysis, manuscript writing and editing.

G.R.R.- Pilot testing, data collection and analysis, manuscript writing and editing.

J.W.H-L.-Project administration, pilot testing, data collection and analysis, statistical analysis, manuscript writing and editing

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Science Foundation (Award #2225135).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data sets generated and analyzed during the current study are available in the Open Science Framework Repository (DOI 10.17605/OSF.IO/SVN32).*

References

Fan

Wang

Song

, et al. Research status and emerging trends in virtual reality rehabilitation: Bibliometric and knowledge graph study. JMIR Serious Games 2023; 11: e41091. https://doi.org/10.2196/41091

Mekbib

Debeli

Zhang

, et al. A novel fully immersive virtual reality environment for upper extremity rehabilitation in patients with stroke. Annals of the New York Academy of Sciences 2021; 1493(1): 75–89. https://doi.org/10.1111/nyas.14554

Cucinella

de Winter

Grauwmeijer

, et al. Towards personalized immersive virtual reality neurorehabilitation: a human-centered design. Journal of NeuroEngineering and Rehabilitation 2025; 22(1): 7. https://doi.org/10.1186/s12984-024-01489-5

Parker

Andreasen

Ricks

, et al. Comparison of brain activation and functional outcomes between physical and virtual reality box and block test: a case study. Disability and Rehabilitation: Assistive Technology 2024; 19(2): 273–280. https://doi.org/10.1080/17483107.2022.2085334

Hadjipanayi

Banakou

Michael-Grigoriou

. Virtual reality exergames for enhancing engagement in stroke rehabilitation: A narrative review. Heliyon 2024; 10(18): e37581. https://doi.org/10.1016/j.heliyon.2024.e37581

Bressler

Merk

Gohlke

, et al. A virtual reality serious game for the rehabilitation of hand and finger function: iterative development and suitability study. JMIR Serious Games 2024; 12: e54193. https://doi.org/10.2196/54193

Oliveira

Marques

Amorim

, et al. Beyond the prototype: Extended use of a virtual reality supermarket in a rehabilitation center. IEEE, 2024, pp. 1–8.

Mani Bharathi

Manimegalai

George

, et al. A systematic review of techniques and clinical evidence to adopt virtual reality in post-stroke upper limb rehabilitation. Virtual Reality 2024; 28(4): 172. https://doi.org/10.1007/s10055-024-01065-1

Demolder

Molina

Hammond

IIIFL

, et al. Recent advances in wearable biosensing gloves and sensory feedback biosystems for enhancing rehabilitation, prostheses, healthcare, and virtual reality. Biosensors and Bioelectronics 2021; 190: 113443. https://doi.org/10.1016/j.bios.2021.113443

10.

Amin

Waris

Syed

, et al. Effectiveness of immersive virtual reality-based hand rehabilitation games for improving hand motor functions in subacute stroke patients. IEEE Transactions on Neural Systems and Rehabilitation Engineering 2024; 32: 2060–2069. https://doi.org/10.1109/TNSRE.2024.3405852

11.

Demeco

Zola

Frizziero

, et al. Immersive virtual reality in post-stroke rehabilitation: a systematic review. Sensors 2023; 23(3): 1712. https://doi.org/10.3390/s23031712

12.

Alashram

Padua

Annino

. Virtual reality for balance and mobility rehabilitation following traumatic brain injury: A systematic review of randomized controlled trials. Journal of clinical neuroscience 2022; 105: 115–121. https://doi.org/10.1016/j.jocn.2022.09.012

13.

Ehioghae

Montoya

Keshav

, et al. Effectiveness of virtual reality–based rehabilitation interventions in improving postoperative outcomes for orthopedic surgery patients. Current Pain and Headache Reports 2024; 28(1): 37–45. https://doi.org/10.1007/s11916-023-01192-5

14.

Feigl

Roth

Gradl

, et al. Sick moves! motion parameters as indicators of simulator sickness. IEEE transactions on visualization and computer graphics 2019; 25(11): 3146–3157. https://doi.org/10.1109/TVCG.2019.2932224

15.

Conner

Freeman

Jones

, et al. Virtual reality induced symptoms and effects: concerns, causes, assessment & mitigation. MDPI, 2022, pp. 130–146.

16.

Laessoe

Abrahamsen

Zepernick

, et al. Motion sickness and cybersickness–Sensory mismatch. Physiology & Behavior 2023; 258: 114015. https://doi.org/10.1016/j.physbeh.2022.114015

17.

Winter

Kern

Gall

, et al. Immersive virtual reality during gait rehabilitation increases walking speed and motivation: a usability evaluation with healthy participants and patients with multiple sclerosis and stroke. Journal of neuroengineering and rehabilitation 2021; 18(1): 68. https://doi.org/10.1186/s12984-021-00848-w

18.

De Keersmaecker

Van Bladel

Zaccardi

, et al. Virtual reality—enhanced walking in people post-stroke: effect of optic flow speed and level of immersion on the gait biomechanics. Journal of NeuroEngineering and Rehabilitation 2023; 20(1): 124. https://doi.org/10.1186/s12984-023-01254-0

19.

Schalbetter

Grêt-Regamey

Gutscher

, et al. High-fidelity immersive virtual reality environments for gait rehabilitation exergames. Frontiers in Virtual Reality 2025; 5: 1502802. https://doi.org/10.3389/frvir.2024.1502802

20.

Lewis

Waltz

Scelina

, et al. Gait patterns during overground and virtual omnidirectional treadmill walking. Journal of NeuroEngineering and Rehabilitation 2024; 21(1): 29. https://doi.org/10.1186/s12984-023-01286-6

21.

Lheureux

Lebleu

Frisque

, et al. Immersive virtual reality to restore natural long-range autocorrelations in Parkinson’s disease patients’ gait during treadmill walking. Frontiers in physiology 2020; 11: 572063. https://doi.org/10.3389/fphys.2020.572063

22.

Soon

Lee

Lau

, et al. Potential of the omnidirectional walking platform with virtual reality as a rehabilitation tool. Journal of Rehabilitation and Assistive Technologies Engineering 2023; 10: 20556683231161576. https://doi.org/10.1177/20556683231161574

23.

Wehden

Reer

Janzik

, et al. The slippery path to total presence: How omnidirectional virtual reality treadmills influence the gaming experience. Media and Communication 2021; 9(1): 5–16. https://doi.org/10.17645/mac.v9i1.3170

24.

Bashir

De Regt

Jones

. Comparing a friction-based uni-directional treadmill and a slip-style omni-directional treadmill on first-time HMD-VR user task performance, cybersickness, postural sway, posture angle, ease of use, enjoyment, and effort. International Journal of Human-Computer Studies 2023; 179: 103101. https://doi.org/10.1016/j.ijhcs.2023.103101

25.

Soni

Lamontagne

. Characterization of speed adaptation while walking on an omnidirectional treadmill. Journal of neuroengineering and rehabilitation 2020; 17(1): 153. https://doi.org/10.1186/s12984-020-00787-y

26.

Jochymczyk-Woźniak

Nowakowska

Polechoński

, et al. Physiological Gait versus Gait in VR on multidirectional treadmill—comparative analysis. Medicina 2019; 55(9): 517. https://doi.org/10.3390/medicina55090517

27.

Luciano

Minetti

Pavei

. Metabolic cost and mechanical work of walking in a virtual reality emulator. European Journal of Applied Physiology 2024; 124(3): 783–792. https://doi.org/10.1007/s00421-023-05306-0

28.

Riley

Paolini

Della Croce

, et al. A kinematic and kinetic comparison of overground and treadmill walking in healthy subjects. Gait & posture 2007; 26(1): 17–24. https://doi.org/10.1016/j.gaitpost.2006.07.003

29.

Hollman

Brey

Robb

, et al. Spatiotemporal gait deviations in a virtual reality environment. Gait & posture 2006; 23(4): 441–444. https://doi.org/10.1016/j.gaitpost.2005.05.005

30.

Hollman

Brey

Bang

, et al. Does walking in a virtual environment induce unstable gait? An examination of vertical ground reaction forces. Gait & posture 2007; 26(2): 289–294. https://doi.org/10.1016/j.gaitpost.2006.09.075

31.

Osaba

Martelli

Prado

, et al. Age-related differences in gait adaptations during overground walking with and without visual perturbations using a virtual reality headset. Scientific Reports 2020; 10(1): 15376. https://doi.org/10.1038/s41598-020-72408-6

32.

Lohman

Turchet

. Evaluating cybersickness of walking on an omnidirectional treadmill in virtual reality. IEEE Transactions on Human-Machine Systems 2022; 52(4): 613–623. https://doi.org/10.1109/thms.2022.3175407

33.

Faul

Erdfelder

Lang

, et al. G* Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior research methods 2007; 39(2): 175–191. https://doi.org/10.3758/bf03193146

34.

McGibbon

Krebs

. Discriminating age and disability effects in locomotion: neuromuscular adaptations in musculoskeletal pathology. Journal of applied physiology 2004; 96(1): 149–160. https://doi.org/10.1152/japplphysiol.00422.2003

35.

Kibushi

. Muscle coordination patterns in regulation of medial gastrocnemius activation during walking. Human Movement Science 2023; 90: 103116. https://doi.org/10.1016/j.humov.2023.103116

36.

Skog

Handel

Nilsson

, et al. Zero-velocity detection—An algorithm evaluation. IEEE transactions on biomedical engineering 2010; 57(11): 2657–2666. https://doi.org/10.1109/TBME.2010.2060723

37.

Abdi

. Bonferroni and Šidák corrections for multiple comparisons. Encyclopedia of measurement and statistics 2007; 3(01): 2007.

38.

Souman

Giordano

Schwaiger

, et al. CyberWalk: Enabling unconstrained omnidirectional walking through virtual environments. ACM Transactions on Applied Perception (TAP) 2011; 8(4): 1–22. https://doi.org/10.1145/2043603.2043607

39.

Egerton

Danoudis

Huxham

, et al. Central gait control mechanisms and the stride length–cadence relationship. Gait & posture 2011; 34(2): 178–182. https://doi.org/10.1016/j.gaitpost.2011.04.006

40.

Janeh

Langbehn

Steinicke

, et al. Walking in virtual reality: Effects of manipulated visual self-motion on walking biomechanics. ACM Transactions on Applied Perception (TAP) 2017; 14(2): 1–15. https://doi.org/10.1145/3022731

41.

Hollman

Watkins

Imhoff

, et al. A comparison of variability in spatiotemporal gait parameters between treadmill and overground walking conditions. Gait & posture 2016; 43: 204–209. https://doi.org/10.1016/j.gaitpost.2015.09.024

42.

Ogawa

Kawashima

Ogata

, et al. Predictive control of ankle stiffness at heel contact is a key element of locomotor adaptation during split-belt treadmill walking in humans. Journal of neurophysiology 2014; 111(4): 722–732. https://doi.org/10.1152/jn.00497.2012

43.

Ibala

Coupaud

Kerr

. Comparison of the muscle pattern variability during treadmill walking (fixed and self-pace) and overground walking of able-bodied adults. Journal of Annals of Bioengineering 2019; 1: 1–11.

44.

Ochoa

Sternad

Hogan

. Treadmill vs. overground walking: different response to physical interaction. Journal of neurophysiology 2017; 118(4): 2089–2102. https://doi.org/10.1152/jn.00176.2017

45.

Padmanaban

Konrad

Stramer

, et al. Optimizing virtual reality for all users through gaze-contingent and adaptive focus displays. Proceedings of the National Academy of Sciences 2017; 114(9): 2183–2188. https://doi.org/10.1073/pnas.1617251114