Meeting Abstract
Robot locomotion on uneven terrain is typically assumed to require complex sensing, control and planning. However, discoveries of the role of mechanics and nonlinear dynamics in running animals and legged robots indicate that stabilization and performance increases can be facilitated via leveraging of non-locomotor structures. Here we examine how an open-loop controlled tail affects performance of a legged robot on uneven terrain. We constructed a RHex-type quadruped robot (L=27cm, m=2.8kg) with compliant C-legs (d=8cm) and a tail (L=20cm, m=0.4kg). Each leg was controlled via a cascaded PID position-velocity control system, with setpoints determined by the duty factor and phase lag of a chosen gait. A landscape consisting of a Gaussian height distribution of 128 blocks (h=0-10cm, w=5cm) generated failures in no-tail robot by either catching a leg or trapping the robot on its belly. We first tested two tail behaviors which we hypothesized would improve performance. The first maintained a constant angle with the body, essentially adding a fifth point of support, and the second oscillated the tail periodically, resulting in intermittent ground contact. In all tests, stability (the average summed roll and pitch of the robot), energy cost (average current draw), and success probability (full transit across the testbed) were measured. In 106 total trials, both tail strategies improved the success probability from 60% (no tail) to 90-100%. Constant angle led to stable locomotion (average of 5.5°) but with high energy cost (0.7A), whereas tapping displayed higher instability (9°) but lower energy costs (0.5A). A “gentle tap” strategy which combined both behaviors demonstrated high success probability (100%), good stability (5.7°), and low energy cost (0.35A), thereby improving locomotor performance with minimal control additions.
