Walking over complex terrain such as a rocky trail or crumbling sidewalk demands precision to avoid tripping, coordination and stability to maintain balance, and endurance to ward off fatigue.  Nevertheless, locomotion would be far more hazardous and exhausting if the motor system had to control all the intricate details of the movement and supply all the forces needed to move the body.  Fortunately, the motor system is opportunistic - it takes advantage of what is available for free from the environment to simplify coordination of the many segments of the body and to move with a minimal amount of self-generated, muscular force.  This understanding led us to new insights into the role that vision plays in guiding locomotion and the control strategies that walkers employ to negotiate extended stretches of irregular terrain.

To bring the task of walking over complex terrain into the laboratory, we developed a novel apparatus that allowed us to manipulate the visual environment and record behavior in ways that are not possible in the real world.  The setup makes use of augmented reality to project virtual terrain onto the floor, which gives subjects a compelling impression of realistic, three-dimensional terrain when viewed while wearing special shutter glasses (Image 1).  As subjects walk, their movements are tracked by a full-body motion capture system.  By integrating the mocap system with the projection system, we can manipulate the appearance of terrain in ways that are synchronized with subjects' movements.

Key finding #1: Walkers exploit the physical dynamics of the body even in complex terrain.  Previous research established that when humans walk over flat, obstacle-free terrain, they exploit the pendulum-like structure of the body to the benefit of energetic efficiency and dynamic stability.  In the presence of more complex terrain, however, the need to avoid obstacles and step on safe footholds entails neuromuscular intervention that could interfere with this natural trajectory.  Nevertheless, we found that walkers do not give up on exploiting external forces and passive stability.  Even when obstacles are present, they favor footholds that allow the body's center-of-mass trajectory to approximate an inverted pendulum.  When multiple foothold options are available, they often prefer the one that allows them to minimally interfere with the body's ballistic movement.  Even when the intended foothold suddenly changes location, walkers prefer to adapt by tailoring the global, pendular mechanics of the body (if time permits) rather than by altering the trajectory of the foot mid-swing.

Key finding #2: Visual information about a potential foothold is most useful when it is sampled 1 to 1 1/2 steps in advance.  Although vision plays an integral role in adapting the ongoing gait cycle to the environment, walkers do not need to continually look at the upcoming terrain.  We found that even a brief glance at the next foot target is sufficient for accurate stepping, as long as that glance occurs at just the right time.  For example, for terrain with clearly defined target footholds, visual information is primarily utilized during the preceding step.  We refer to this as the critical control phase (Image 2).  When visual look-ahead is limited such that walkers cannot see the upcoming terrain until after the beginning of the critical control phase, performance degrades.  Likewise, performance is at most weakly affected when visual information about the locations of upcoming targets is made unavailable before or after the critical control phase.  The preference to sample visual information during the critical control phase has its roots in the physical dynamics of walking, for this is when the determinants of the body's ballistic trajectory are under the walker's control.

Key finding #3: In 3d terrain, walkers use information from farther in advance.  Although previous research examined the visual control of stepping over single obstacles in isolation, few studies have examined walking over extended stretches of three-dimensional terrain.  The 3d floor-projection setup allowed us to manipulate when terrain was visible to walkers as they moved to determine how the critical phase for visual control varies with terrain type and complexity.  Interestingly, when raised obstacles are present, walkers benefit from being able to see a bit farther ahead compared to when obstacles are flat.  By using visual information from farther ahead, walkers are not only better able to step over obstacles but also better able to avoid having to step over obstacles in the first place. 

Broader impacts: Locomotion is a fundamental aspect of human behavior and is integral to many daily human activities.  The results of this project inform our understanding of the control of locomotion in healthy adults, which in turn helps medical doctors and other health-care practitioners anticipate and manage the consequences of visual and motor impairments on mobility.  The results also provide a source of ideas for roboticists who are exploring bio-inspired solutions to developing bipedal robots that can walk over real-world terrain. 

Last Modified: 01/06/2020
Modified by: Brett R Fajen