Robots and animals transition from less to more favorable modes to traverse obstacles

Legged robots still struggle to robustly traverse obstacles in complex terrain using physical interaction, which animals do at ease. Recent studies revealed that cockroaches use body-beam interaction to traverse grass-like beam obstacles, probabilistically transitioning between locomotor modes. To understand what governs the direction of locomotor transitions between modes, we developed and studied a robophysical model traversing beams of variable stiffness. We discovered that, when viewed on a potential energy landscape resulting from body-beam interaction, different locomotor modes (system state trajectories) were attracted to different local minima basins. Regardless of beam stiffness, the system was always more likely to transition from less to more favorable modes (i.e., from lower to higher energy local minima) on the landscape, as long as kinetic energy fluctuation from oscillatory self-propulsion is sufficient to overcome the potential energy barrier in between. We made similar observations in cockroaches, despite their active behaviors. These physical principles are surprisingly similar to microscopic systems (e.g., protein-folding transitions on a free energy landscape). Our study is a step in establishing energy landscapes for locomotor transitions in complex terrain.