Meeting Abstract
The knowledge biologists have uncovered studying animal systems has and continues to improve robotic systems through a program of bioinspired design. We extend this program by transferring dog control strategies to a robotic system and measuring performance changes. When dogs move at walking speeds on uneven terrain their walk becomes more variable, taking on some characteristics of a trotting gait. This behaviour is consistent with the recovery of the dogs gait being optimized for static stability, but could also be a result of the recovery of the gait being optimal for some other factors (such as energetics, angular stability), or a result of symmetries or constraints on the animal. In an effort to better understand the causes of this dynamic in the dog walk (and potentially improve the control of robotic legged systems) we employ a physical model, transferring a putative mathematical representation of this control strategy to the RHex robot with the aim of inducing analogous behaviour. We measure the degree to which this changes the angular stability of the robot, as well as measuring the change in cost of transport with the aim of trying to resolve which factors benefit most from this change in control strategy. We compare three different control strategies, the standard RHex feed-forward controller, a controller with feedback, and a controller with feedback and recovery to the standard walk structured so that on rough ground the gait takes on some characteristics of a trot.
