Market-based task assignment strategies for multi-agent systems deployed for bushfire fighting

This paper studies the task assignment strategies for multi-agent systems to cooperatively accomplish a set of tasks while achieving a team objective that give near-optimal final assignments, in the context of simulated bushfire fighting scenario. The purpose of this research is to develop efficient strategies to employ multiple robots to cooperate to extinguish a bushfire with multiple fire fronts by delivering sufficient extinguishing agents to each fire fronts and for each agent to replenish its resources between every assigned fire front. We address the problem by extending the existing market-based auction algorithm to incorporate the use of a bushfire prediction model. We approach this problem with saving the properties and populations as the main objective. However, this objective does not make the property location a target for the robots nor the entire wildfire boundary being selected as targets. Instead, we propose a target selection model that determines the rendezvous point of the agents and the critical fire fronts which poses the most threats to property or human life. The complexity of the problem is mainly due to the dynamic nature of the bushfire spreading. However, this can be taken into account with a highly reliable bushfire prediction model that considers the majority of the significant factors that affects the spreading of the fires. The auction algorithm auctions the destinations for the agents which in fact are the critical points at which the agents rendezvous the fire fronts. The modifications to the standard auction algorithm are also presented.