DisCo: distributed contact-rich trajectory optimization for forceful multi-robot collaboration
摘要
We present DisCo, a distributed algorithm for contact-rich, multi-robot tasks, which enables a team of robots to optimize a time sequence of forces to manipulate objects or locomote via contact interactions with their environment to accomplish dexterous robotics tasks such as collaborative manipulation, robot team sports, and modular robot locomotion. We leverage a fully-distributed variant of the Alternating Direction Method of Multipliers (ADMM) to decompose the global trajectory optimization problem into smaller single-robot, contact-implicit problems. Each robot only computes its own contact forces and relevant contact-switching events from its local optimization problem, enabling three key advantages: (i) significant generality to a broad range of contact-rich problems, (ii) superior task success rates, and (iii) much faster computation times, compared to existing methods. Notably, the local problems solved by each robot are significantly less challenging than a centralized problem with all robots’ contact forces and switching events, improving the computational efficiency, while also preserving the privacy of some aspects of each robot’s operation. We demonstrate the effectiveness of our algorithm across a wide range of problems including collaborative manipulation, multi-robot team sports scenarios, and in modular-robot locomotion, where DisCo achieves 3x higher success rates over a baseline success rate of