Tasks Allocation to Agents using Market-Based Techniques

We consider the problem of allocating autonomous agents to perform multiple tasks, de ned by having geographic location, priority, and a time frame. We formulate the problem as a Multiple Traveling Salesmen Problem (MTSP) With Bene ts, which is a generalization of the well known Traveling Salesman Problem (TSP). Similar to the TSP, optimal solutions to the MTSP are considered NP hard, but there are many approximate solutions with reduced complexity. We propose a market based technique, where each task is represented by an agent that o ers a time varying \bene t" to be assigned. Each autonomous agent o ers to perform a task based on the cost of traveling to the location of that task and the estimated time to perform it. Tasks are assigned to agents, which can later trade tasks among themselves to achieve a required global goal. In this work we present preliminary results for a simpli ed version of the problem, which corresponds to the common MTSP, where all tasks have the same constant bene t, and all of the tasks have to be performed. A comparison of results vs. optimal solutions obtained for simple cases and with genetic algorithm solutions for more complicated cases is presented, as well as the scalability of this approach. Future directions for this approach are discussed.