Distributed Task Allocation in Dynamic Multi-agent System

Abstract

Distributed task allocation has been the hot research topic from the last few years. It is the heart of multi-agent systems. In multi-agent system, the agents coordinate and cooperate with other agents to accomplish the complex task which cannot be completed by an individual agent. Here, a distributed task allocation approach is proposed in constrained cooperative multi-agent environment (dynamic, real-time and uncertain). Agent allocates the task to multiple agents by considering the spatial, temporal and communicational constraints of the environment. The proposed approach considers the negotiation-based task allocation approach where the main agent announces the task and then other agents sends their respective bids for the received task. Best bid is chosen from all the received bids and then task is allocated to winning agent or group of agents. The main objective is to minimize the waiting time for a task to be accomplished and the number of messages transferred among agents for task allocation process. Furthermore, due to uncertainty of dynamic environment where the environment gets evolved at any point of time and plan gets failed, a re-planning algorithm is proposed which enables the agents to re-coordinate their plans when environment problem avoid it to fulfill them. The proposed approach is applied to the fire-fighting multi-agent environment where the allocation of firebrigade agents is done to extinguish the fire in an efficient and effective manner. The approach is simulated in a multi-agent framework JADE and the result shows that the proposed approach requires less number of messages and less waiting time for the successful task allocation.