An Augmented Lagrangian Method in Distributed Dynamic Optimization Based on Approximate Neighbor Dynamics

In this paper, a Lagrangian decomposition scheme for the agent based distributed dynamic optimization of coupled nonlinear continuous-time systems is presented. In contrast to existing decomposition schemes, each agent is augmented with approximate dynamics of the coupled neighbor agents, thus enabling the agent to anticipate the dynamic behavior of his neighbors. The performance of the presented decomposition scheme is compared to a standard decomposition scheme by simulation results for a cooperative payload transport by a team of physically coupled robots.