Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion

Abstract

In this paper, we propose a framework to build a memory of motion for warm-starting an optimal control solver for the locomotion task of a humanoid robot. We use HPP Loco3D, a versatile locomotion planner, to generate offline a set of dynamically consistent whole-body trajectory to be stored as the memory of motion. The learning problem is formulated as a regression problem to predict a single-step motion given the desired contact locations, which is used as a building block for producing multi-step motions. The predicted motion is then used as a warm-start for the fast optimal control solver Crocoddyl. We have shown that the approach manages to reduce the required number of iterations to reach the convergence from ∼9.5 to only ∼3.0 iterations for the singlestep motion and from ∼6.2 to ∼4.5 iterations for the multi-step motion, while maintaining the solution's quality.

Receding-Horizon Perceptive Trajectory Optimization for Dynamic Legged Locomotion with Learned Initialization

“…One increasingly popular strategy for online planning is to produce high-quality initial guesses by leveraging offline experience, so that only minor computation is needed to adapt to a scenario seen at runtime. Many earlier examples of this approach focused on manipulators [2], leading to later studies on more dynamic systems [3], [4], [5]. Recent works have considered more complex variants that may be needed when the relevant solution space is multimodal [6], [7].…”

“…Guided Trajectory Learning uses consensus optimization to learn a function approximator that only reconstructs feasible motions [32]. In [4], a "memory of motion" of dynamic, collision-free locomotion has been generated using the HPP Loco3D planner [33] for the Crocoddyl solver [10].…”