Decentralized Cooperative Trajectory Optimization for UAVs with Coupling Constraints

Motivated by recent research on cooperative UAVs, this paper introduces a new decentralized trajectory optimization approach for systems with independent dynamics but coupled constraints. The primary objective is to improve the performance of the entire fleet by solving local optimization problems and without reproducing the global optimization problem for each agent. To achieve cooperation, the approach exploits the sparse structure of active couplings that is inherent in the trajectory optimization. This enables each local optimization to use a low-order parameterization of the other agents' states, thereby facilitating negotiation while keeping the problem size small. The key features of this approach include: (a) no central negotiator is required; and (b) it maintains feasibility over the iterations, so the algorithm can be stopped at any time. Furthermore, the local optimizations are shown to always decrease the overall cost. Simulation results are presented to compare the distributed, centralized and other (non-cooperative) decentralized approaches in terms of both computation and performance