Locomotor circumvention strategies in response to static pedestrians in a virtual and physical environment

Highlights

• Obstacle circumvention in virtual reality elicits larger obstacle clearances.

• Slower walking speeds are consistently observed in virtual reality.

• Characteristics of circumvention strategies are preserved in virtual reality.

• Variability in circumvention outcomes is similar in virtual and physical environments.

Abstract

Background

Circumvention of pedestrians is an essential requirement of community ambulation and can be challenging to reproduce in laboratory or clinical settings. Virtual reality (VR) is a powerful tool that allows investigations, assessments or training of such tasks under ecological but controlled conditions. The extent to which current VR technologies can elicit responses similar to those observed in the physical world, however, remains to be determined.

Research questions

(1) To what extent does the circumvention of static pedestrians in VR differ from that observed in the physical environment (PE)? and; (2) To what extent does the inter-trial variability of obstacle circumvention outcomes differ in VR vs. the PE?

Methods

Healthy young participants (n = 13) were assessed while walking and avoiding a collision with an interferer that stood either at 3.0 and 3.5 m from the participant's starting position (experimental trials) or that exited to the side (catch trials). The task was performed in the PE and VE, in a random order. A female collaborator acted as interferer in the PE and her kinematics was used to create the avatar used in the VE.

Results

Compared to the PE, the circumvention of a static pedestrian in VR was characterized by larger obstacle clearances and slower walking speeds. Characteristics of circumvention strategy such as the preferred side of circumvention, response to obstacle position and pattern of speed adaptation were similar between VR and the PE. Inter-trial variability for the different outcomes were also similar between the two environments.

Significance

Differences in obstacle clearance and speed indicate the use of “safer” circumvention strategies in VR. However, the patterns of locomotor adaptation that were largely similar between the two environments which suggests that VR is a valuable tool to study, assess and possibly train complex locomotor tasks such as obstacle avoidance.

Introduction

Obstacle circumvention is essential for a safe and independent ambulation in community settings [1]. It requires one to assess and update information of self-motion in relation to objects and other people that may obstruct the travel path. This information is primarily obtained by the visual system through the use of several depth perception visual cues and by measuring changes in one's gaze-movement angle (i.e. bearing angle [2]). In the presence of obstacles, pedestrians estimate the point in time when a collision would occur [2] and coordinate adjustments of walking trajectory and speed to ensure obstacle clearance [3].

The distance maintained from the obstacle during its circumvention is thought to reflect the safety margin deemed necessary by the pedestrian to successfully avoid an obstacle [4]. Studies have shown that while this clearance remains constant in response to factors such as walking speeds [5], [6], it is enlarged under conditions of increased cognitive load (e.g. dual tasking [7]) and in the presence of sensory distractors (e.g. auditory sounds [8]). Recent studies further suggest that the nature of the obstacle also influences obstacle circumvention. For instance, larger clearances were observed when pedestrians crossed an aperture delimited by two confederates as opposed to two poles [9]. A study from our laboratory further showed that young adults use smaller obstacle clearances when avoiding moving pedestrians vs. cylinder in a virtual environment (VE [10]).

Replicating experimentally obstacle circumvention strategies in response to human obstacles in research and clinical setting can be challenging. Fortunately, the appropriate conditions can be reproduced in controlled, safe and ecological environments using virtual reality (VR). However, as VR simulations are not yet capable of delivering simulations that accurately embody all sensory systems, the extent to which they evoke natural motor responses can be disputed [11]. Distance reports are consistently underestimated in VEs [12], [13] and field of view (FOV) restrictions, as often experienced when using helmet mounted displays (HMDs), have been shown to induce slower walking speeds [13], [14] and a higher reliance on head movements to visually scan the environment [12]. In fact, studies which examined obstacle circumvention in response to static virtual cylinders or boxes reported the use of “safer” avoidance strategies, with subjects adopting larger obstacle clearances and slower walking speeds in the VE compared to the PE [5], [15], [16]. Although to a smaller extent, these differences were also observed as participants circumvented animate (another person) and inanimate (box) static obstacles while walking in a VE projected with a CAVE system [16]. Whether such findings would also be extended to the circumvention of a static pedestrian while immersed in a low-cost HMD-based VR system, which allow the user to move freely over a larger volume, remains to be determined.

In addition, studies performed in the PE have described stereotypical walking trajectory adaptations both during goal directed locomotion [17] and obstacle circumvention [18]. However, since these adaptations largely depend on spatial information acquired through vision [11], it is unclear whether locomotion in VE exhibits similar variability when compared to the PE. Also, while variability of locomotor performance in VEs was examined before, existing reports were largely limited to spatiotemporal distance factors as outcomes reflecting gait stability during steady gait [19], [20] and information about obstacle circumvention is lacking. Lastly, studies that compared obstacle circumvention in HMD-based VEs and PEs were conducted more than a decade ago and recent developments of virtual reality technologies now allows for higher quality graphics, enhanced real-time rendering, as well as HMDs with larger FOV and lighter weight, which highlights the importance of revisiting this topic.

This study was carried out as part of a larger project where obstacle circumvention strategies in response to human obstacles, either stationary or in motion, were examined in the VE and PE. In this specific study, we estimated the extent to which the circumvention of a static human obstacle differed in the VE vs. PE in healthy young adults. We further examined the inter-trial variability of obstacle circumvention outcomes in both environments. We hypothesized that due to a combination of perceptual and cognitive factors associated with performing the task in a VE, participants would maintain larger obstacle clearances and adopt slower walking speeds in the VE, which in sum is characterized as a “safer” circumvention strategy. It was further hypothesized that obstacle circumvention outcomes would display a larger inter-trial variability in the VE as compared to the PE.