Coordinated Multi-Agent Motion Planning Under Realistic Constraints

Abstract

Considered is a class of cooperative control problems that has a special affine characterization. Included in this class of multi-agent problems are the so called radar deception problem, formation keeping and formation reconfiguration. An intrinsic geometric formulation of the associated constraints unifies this class of problems and it is the first time such a generalization has been presented. Based on this geometric formulation, a real-time motion planning algorithm is proposed to generate dynamically feasible reference trajectories for the class. The proposed approach explicitly considers actuator and operating constraints of the individual agents and constrained dynamics are derived intrinsically for the multi-agent system which makes these constraints transparent. Deriving the constrained dynamics eliminates the need for nonlinear programming to account for the system constraints, making the approach amenable to real-time control. Explicit consideration of actuator and operating limitations and nonholonomic constraints in the design of the reference trajectories addresses the important issue of dynamic feasibility. The motion planning algorithm developed here is verified through simulations for the radar deception, rigid formation keeping and formation reconfiguration problems. A key objective of this study is to advocate a change in paradigm in the approach to formation control by addressing the key issues of dynamic feasibility and computational complexity. The other important contributions of this study are: Unifying formulation of constrained dynamics for a class of problems in formation control through the intrinsic geometry of their nonholonomic and holonomic constraints; Deriving these constrained dynamics in any choice of frame that can even be coordinate free; Explicit consideration of actuator and operating limits in formation control to design dynamically feasible reference trajectories and Developing a real-time, distributed, scalable motion planning algorithm applicable to a class of autonomous multi-agent systems in formation control.