Formation adaptation in obstacle-cluttered environments via MPC-based trajectory planning
Chen, Yuda; Li, Zhongkui
Sci China Inf Sci, 2024, 67(7): 174201
Formation is a typical collective behavior in multiagent systems utilized in areas such as inspection, surveillance, and transportation. Commonly-used approaches to achieve formation include virtual structure, consensus-based, affine formation, and optimization-based methods. While these feedback control methods are known for their robustness and low computation requirements, they often neglect the challenges posed by obstacle-cluttered environments where the agents need to adapt their formation structure to safely navigate through narrow passages or around corners. Recent studies have introduced the concept of adopting trajectory planning to achieve formation while avoiding obstacle collisions. This involves reformulating the planning process as a numerical optimization problem, enforcing either soft or hard constraints to prevent collisions with obstacles and between agents. However, there are two notable issues with existing approaches: (1) the induced constrained optimization might not always be feasible, meaning that there is no theoretical guarantee for the safety of the underlying agents regarding both obstacle avoidance and collision prevention among agents; (2) interagent communication in the formation, e.g., the line-of-sight collision between agents, has not been adequately addressed.