Efficient Multicontact Pattern Generation With Sequential Convex Approximations of the Centroidal Dynamics

This paper investigates the problem of efficient computation physically consistent multi-contact behaviors. Recent work showed that under mild assumptions, could be decomposed into simpler kinematic and centroidal dynamic optimization problems. Based on this approach, we propose a general convex relaxation dynamics leading to two computationally algorithms based iterative resolutions second order cone programs. They optimize trajectories, contact forces and, importantly, timing motions. We include approach in kino-dynamic method generate full-body movements. Finally, is embedded mixed-integer solver further find dynamically sequences. Extensive numerical experiments demonstrate computational efficiency suggesting it used fast receding horizon control loop. Executions planned motions simulated humanoids quadrupeds real quadruped robot show quality optimized