Meeting Abstract

SICB+    A snake robot to study force generation and control of limbless locomotion in 3-D terrain Ramesh, D*; Fu, Q; Li, C; Johns Hopkins University dramesh6@jhu.edu http://li.me.jhu.edu

Generalist snakes use body-terrain contact forces to transition between different strategies swiftly and stably to traverse complex 3-D terrain. Our recent animal observations suggested that snakes use contact force sensing to control body bending to generate propulsive forces. In our previous study, we developed a robotic physical model that used a pressure sensor array, with 8 and 4 sensors for each yaw and pitch segment, respectively, to detect contact forces. However, the previous robot had moved poorly over obstacles due to its heavy weight and limited range of motion from the total of 72 sensors and suffered a low sampling frequency. Here, we developed a refined robot to overcome these limitations. We replaced the sensor array with a single sheet sensor to reduce the number of sensors on the robot. The new 12-segment robot has 36 sensors along the entire body with 3 sensors on each segment, achieving a sensing frequency of 30 Hz. Each sensor can detect normal contact forces up to 6.5 N and 3.25 times the segment weight. Each segment has a 3-D printed soft shell that can conform to unknown terrain to distribute contact force more evenly on the sensor. A mesh sleeve covers the entire robot to protect the sensors and reduce frictional drag. As a first step to demonstrate the usefulness of force sensing, we demonstrated the robot traversing a half-cylindrical obstacle by propagating a predefined vertical bending shape that conformed to obstacle geometry down its body. We observed large forces on the body segments in contact with the frontal side of the obstacle, resulting in net forward propulsion to overcome frictional drag. We are currently exploring force feedback control strategies to understand how to use force information to control body bending to better conform to and push against the terrain to move effectively. Also see related work in another talk by Fu & Li.