Trajectory Optimization Methods for Energy Efficient Gait Transitions on Multi-Modal Robots

For robots that achieve locomotion through the movement of articulation joints, having a set of gaits can allow them to navigate variable terrain. Effectively transitioning between these gaits on multi-modal robots is important for maintaining locomotive constraints and reducing unnecessary movements during gait switches, which decreases energy usage. While previous works have examined transitions between specific gaits on certain robots, this work aims to propose and experimentally examine general frameworks for transitioning between gaits with a nonlinear, local optimized trajectory over the articulation joints. The proposed frameworks are targeted towards compute-constrained platforms that require efficient and online trajectory generation. We compare the maximum joint acceleration, transition time, cost of transport, and heading change between four transition frameworks: an optimized Bezier function, optimized Linear waypoints, an optimized Spline, and resetting to a neutral position before each gait. These transition strategies were then implemented and experimentally evaluated on a four-wheeled multi-modal rover.