On the optimal spine morphology of rapidly accelerating quadrupeds

Abstract

Animals exploit spine actuation during rapid locomotion, however this has only recently become a focal point in robotics. Roboticists have used a multitude of spine configurations in their platforms but the optimal design for rapid acceleration and deceleration maneuvers is yet to be discovered. In this paper, we endeavour to find this optimal spine morphology by using large-scale Monte Carlo trajectory optimization simulations on long-time-horizon minimum time problems (start and end at rest while travelling a fixed distance of 30 spine lengths). Broad applicability of the results was ensured by generating 100 sets of robot parameters at random from a carefully selected design space, comparing the performance of the rigid, revolute and prismatic spine morphology. Using bootstrapping techniques, it was determined with a 78.8% probability that the prismatic spine morphology was the optimal spine for these long-time-horizon trajectories. These results will serve as a guide for designers of future, agile quadruped robots. Keywords: legged robot, trajectory optimization, optimal control.