Efficient CoM Motion Planning for Quadruped Robots’ Quasi-Static Walking

With the popularity of quadruped robots, the main challenge they must overcome is traversing unstructured environments. Current methods that allow modern robots to traverse challenging terrain are unsuitable for situations at the edge of robot performance, where torque limits and contact forces must be carefully considered. This paper will investigate a way of generating feasible center of mass (CoM) trajectories applicable in such cases. A feasible CoM trajectory is one that the robot can perform considering contact, torque, and reachability constraints. We improve the existing method for finding feasible CoM regions, yielding a thirty times speedup so that it can run under 1 ms. Based on that improvement, we introduce a new iterative CoM planner that sequentially solves prioritized constrained IK and computes feasible regions. That way, we guarantee the satisfaction of contact constraints, torque constraints, and reachability. The planned motion was performed using a whole-body controller. We tested the approach on high-fidelity simulation and on real Solo12 quadruped, achieving the control loop frequency of 1 kHz. The whole codebase has been disclosed on GitHub.