Multi-Agent Task and Motion Planning: An Optimization based Approach

We propose a new algorithm for Multi Agent Task and Motion Planning (TAMP). Our approach builds on the Logic-Geometric Programming framework (LGP) presented in prior work [1, 2]. The presented algorithm plans policies that react to the actions of the other agents, both on the symbolic and the motion level. To this end, we optimize trajectory trees that describe the branchings of optimal motions depending on the other agent actions. The algorithm works in two stages: First, the symbolic policy is optimized using approximate path costs estimated from independent optimization of trajectory pieces. Second, we fix the best symbolic policy and optimize a joint trajectory tree.