Meeting Abstract

13-4   11:15 - 11:30  Robotic snake traversing large obstacles using vertical bending with contact force feedback Fu, Q*; Li, C; Johns Hopkins University; Johns Hopkins University fqiyuan1@jhu.edu http://li.me.jhu.edu/

Snakes can traverse complex 3-D terrain. Previous snake and snake robot studies explored how to control lateral body bending to generate propulsion against vertical structures like vertical pegs and walls on flat surfaces. However, we know little about how to control vertical body bending to push against terrain height variation such as a horizontal ladder, a wedge (Jurestovsky, Usher, Astley, 2021 JEB), and uneven terrain (Fu, Astley, Li, 2021 SICB). Here, we explore this question by testing two hypotheses in a robotic physical model: (1) Vertical bending alone can provide propulsion on terrain with large height variation; (2) Force feedback can improve the performance by maintaining propulsive forces against the environment. The robot has 9 active pitch joints controlled by servo motors and passive wheels with small friction. Although lacking vision, the robot has a force sensing resistor between each wheel and its segment to measure approximate normal terrain reaction forces, which can be used to control its pitch joint. To test the first hypothesis, we challenged the robot to traverse a half cylindrical obstacle as high as 11% of its length while hauling a constant load. By propagating a fixed vertical bending shape that matched the obstacle geometry down the body, the robot traversed with a load as large as 3 times its own frictional drag. To test the second hypothesis, we added obstacles of unknown shapes in front of the half cylinder and compared traversal performance with and without force feedback control. Using force feedback to adapt to terrain without reducing active pushing, the robot better traversed while maintaining good obstacle contact. These observations supported our hypotheses. We are currently exploring how to coordinate multiple body segments to move effectively using force feedback. Also see related work in another talk by Ramesh, Fu, Li.