Modulation of robot orientation by exploiting leg-obstacle collisions through successive choices of gait

An environment can provide different affordances for locomotion based on a locomotor’s actions. Our study aims to explore how multi-legged robots can use various gaits to exploit the affordance of obstacle collisions to negotiate complex environments in different ways.

To allow assessment of obstacle affordances, we represent physical obstacles as a horizontal-plane disturbance field. For a given physical environment, the disturbance field is fixed. However, when a (structured) disturbance field couples with (periodic) locomotor gaits, there emerge stable fixed-point heading directions with substantial basins of attraction.

In this study, we show that the basins around these fixed-point headings depend sensitively on the locomotor gaits. We investigate the transit between these basins as a quadrupedal robot explores its 3-dimensional gait space (each dimension represents the relative phase difference between one pair of legs), and show that these gait-dependent basins allow the robot to take advantage of the repeated leg-obstacle collisions to robustly achieve desired sequences of headings without active steering.