Testing whether potential energy landscape can predict stochastic obstacle traversal

Animals excel at traversing large obstacles in complex terrain by transitioning between locomotor modes, an ability that robots still lack. Our recent study established a quasi-static potential energy landscape approach to locomotor transitions (Othayoth, Thoms, Li, 2020, PNAS). Here we tested whether potential energy landscapes can statistically predict stochastic traversal, specifically, that traversal probability should increase with decreasing potential energy barrier. We developed a dynamic simulation of a simplistic 2-D model system, a self-propelled circular body traversing two horizontal elastic beam obstacles with different stiffness. We assumed each body to be rigid and interacted via collision and continuous contact with no friction. We found that the body had a finite probability to be trapped in an attractive landscape basin in front of two barriers resulting from the two obstacles. Increasing the random force and self-propulsive force increased the probability to escape from this basin and move along trajectories that overcame the lower barrier. These results showed that the potential energy landscapes can predict the probability distribution of stochastic traversal, thus providing a simple, generalizable model for designing robots to traverse complex terrain.