Meeting Abstract

SICB+    Force sensing helps differentiate stiff and flimsy beam obstacles and facilitates traversal Xuan, Q*; Wang, Y; Li, C; Johns Hopkins University qxuan1@jhu.edu

Animals are excellent at traversing complex terrain with cluttered obstacles. For example, the discoid cockroach can traverse cluttered grass-like beams with gaps as narrow as 40% body width (Othayoth, Thoms, Li, 2020, PNAS), using different locomotor modes to cope with beams of different stiffness. When encountering flimsy beams, the animal often pushes across with its body pitched up. When encountering stiff beams, the animal often transitions from pushing to rolling into a gap and maneuvering through, because pushing across requires overcoming a larger potential energy barrier. Before and during this transition, the animal actively adjusts its head, abdomen, and legs to facilitate body rolling (Wang, Othayoth, Li, 2020, SICB). This suggests that the animal can sense resistive forces to differentiate stiff and flimsy beams and switch to the less costly roll mode. Here, we developed a physics model to test the feasibility of this strategy. The animal body was simplified as an ellipsoid, which can sense beam contact forces and actively control its motion. Beams were simplified as rigid plates with torsion springs at the base, with a gap of 83% body width. Noise was added to the sensed forces to emulate the effect of intermittent contact and sensing errors in the real system. The stiffness of each beam estimated from the sensed forces using the model well matched the actual value, with higher accuracy for smaller noise. Using these estimates, the model predicted contact forces and mechanical energetic cost, which was then used to plan and control body motion to traverse with minimal cost. When traversing stiff beams, this strategy rolled the body to traverse, whereas without it the body did not roll and was stuck against the beams. When traversing flimsy beams, this strategy controlled body pitch to avoid flipping over. We demonstrated this strategy in a robotic physical model.