Consensus Based Decoupling in Hybrid Optimization of Cooperative Transportation Planning

Abstract A recently introduced new class of multi-agent transportation planning is considered. Several agents have to cooperate in order to transfer a passive agent from its initial configuration to a goal configuration. Docking events refer to discrete decisions, whereas the individual agents are governed by nonlinear continuous dynamics. Modeling this problem in a hybrid optimal control framework allows computing a solution using a centralized, hierarchical optimization algorithm. Relatively well performing solutions can be found with much lower computational time by decoupling the discrete-event level from the continuous processes. This is accomplished by introducing a feedback controller based on a consensus protocol into the hierarchical optimization structure. The scalability and optimality of this approach is examined for varying information topologies in the consensus protocol and controller gains.