Solution of a Multi-Agent Transport Problem by Hybrid Optimization

As reported in earlier work, methods for hybrid optimal control can be used for optimal path planning of multi-agent systems. This paper considers a multi-vehicle transport scenario in which a hybrid model is used to represent the continuous dynamics of the vehicle motion and the dynamics arising from the docking-events between vehicles and a transported object. The decisions on the vehicles’ heading and speed as well as on the docking sequence are strongly coupled and require a tailored numeric solution. The paper proposes a hierarchical solution strategy with three layers: on the upper layer a discrete event sequencing problem is solved, the middle layer determines docking positions and times, and the lower layer provides optimal vehicle paths. By a combination of graph search and embedded continuous optimal control, the efficiency of obtaining (sub-)optimal solutions is improved. The algorithm is illustrated by a scenario in which three vehicles transport one object.