Group buying algorithm for task assignment of multi-agent load transport system

Abstract

A rotary wing aircraft, so that can be vertical takeoff and landing, it is suitable for the transporting goods in the downtown area and the mountainous area. In particular, a plurality of aircraft are possible to transport a freight, which exceeds the load capacity of each aircraft, in slung load form. As the use of un-manned aircraft is getting diverse, and private usage of UAV increases, the research for the multi-agents task assignment of unmanned cargo transport system is required. This work specifies the task unit as a cargo transport by defining the task tuples, and formulate the multiple cooperative task assignment (MCTA) problem as a utility optimal problem. In addition, the thesis expressed the task assignment of the multiple agents load transport system as the MCTA problem that the initiation of the task is constrained by the critical agents. For the task assignment of the multiple agents load transport system, this thesis proposed a group buying algorithm based on the sequential greedy algorithm. The group buying algorithm has a particular property, which is different from the original sequential greedy algorithm, of being more robust to the impact of the winner’s curse by running the market to maximize the utilities of the critical agents of each task. For the two common problem that can be held in the group buying algorithm, the incentive strategy for resolving the over concentration problem and the additional algorithm for resolving the cross deployment are consid-ered. To obtain the optimal task assignment for the performance analysis, a complete enumeration method that visits all feasible solution and finds the optimal solution is made. Then the cross deployment problem that can occurs when the MCTA is constrained on the initiation of the task by the critical agent is defined mathematically, and proposed an algorithm that can inspect whether the given path is cross-deployed or not. The author conducted the Monte-Carlo Simulation and the analy...