Decentralized Task and Path Planning for Multi-Robot Systems

We consider a multi-robot system with team of collaborative robots and multiple tasks that emerges over time. propose fully decentralized task path planning (DTPP) framework consisting allocation module localized module. Each is modeled as Markov Decision Process (MDP) or Mixed Observed (MOMDP) depending on whether full states partial are observable. The then aims at maximizing the expected pure reward (reward minus cost) robotic team. fuse model into factor graph formulation so can be decentrally solved using max-sum algorithm. robot agent follows optimal policy synthesized for we forward dynamic programming scheme resolves conflicts between agents avoids collisions. proposed demonstrated high fidelity ROS simulations experiments ground robots.