Learning Task Allocation for Multiple Flows in Multi-agent Systems

Task allocation is a key problem for agent to reach cooperation in multi-agent systems. Lately task flows are replacing traditional static tasks, thus real-time dynamic task allocation mechanisms draw more attention. Though scheduling single task flow is well investigated, little work on allocation of multiple task flows has been done. In this paper a distributed and self-adaptable scheduling algorithm based on Q-learning for multiple task flows is proposed. This algorithm can not only adapt to task arrival process on itself, but also fully consider the influence from task flows on other agents. Besides, its distributed property guaranteed that it can be applied to open multi-agent systems with local view. Reinforcement learning makes allocation adapt to task load and node distribution. It is verified that this algorithm improves task throughput, and decreases average execution time per task.