How cockroaches traverse large obstacles with the least effort: insights from robophysical modeling

Cockroaches excel at traversing cluttered large obstacles, outperforming even the best robots, by transitioning across locomotor modes. Previously, we discovered that their body–obstacle interaction in this process results in a potential energy landscape. On this landscape, locomotor transitions are transitions from one attractive basin to another, separated by high potential energy barriers. The animal crosses these barriers via saddle points to traverse obstacles, which incurs the least resistance and effort. But unlike crossing a mountain ridge via its saddle, which can be guided by vision, the cockroach cannot see a potential energy landscape to find its saddle. Here, we provide an answer to this problem. We developed a robotic model of the cockroach traversing grass-like beam obstacles, which can sense obstacle contact forces and torques, and measured these forces and torques over all possible system states. Although they are non-conservative, these forces and torques well approximated the potential energy landscape gradients (conservative forces), allowing estimation of the landscape. But during locomotion, an animal cannot visit all but only states along its trajectory. Curiously, the cockroach’s body oscillates a lot during traversal. Using a validated simulation, we discovered that such large body oscillations while sensing forces and torques allowed estimating the potential energy landscape locally along the trajectory, which is sufficient for finding and crossing a nearby saddle.