Decentralized velocity-aware motion planning for multi-agent coordination

Multi-Agent motion planning is a crucial topic with applications of search and rescue missions, transportation missions, etc. The concerns of motion planning focus on path planning and inter-Agent collision avoidance. Lyapunov-based models present great solutions for the problems. However it is usually empirical to set reasonable parameters for the models. Moreover, agents controlled by the models usually converge to destinations slowly. Heuristic planning algorithms are also mainstream approaches to guide agents without consideration of kinetics. This paper proposes decentralized velocity-Aware planning algorithm and collision avoidance algorithm to serve motion planning of multiple agents. The velocity-Aware algorithm generates paths with acceleration vectors that converge to the predefined destinations. The collision avoidance algorithm will be triggered to protect agents from collisions. The simulation results illustrate the efficiency of our algorithms. Furthermore, compared with the prioritized A∗ algorithm paired with Lyapunov-based low-level control laws, our approach could help agents take much shorter paths and much less time to move to destinations safely.