Multi agent foraging - taking a step further Q-leaming with search

Abstract

The paper discusses a foraging model which accomplishes coordination obliged tasks. This is done through communication techniques and by learning from and about other agents in a confined, previously unseen environment. A new reinforcement learning technique, Q-Learning with search has been proposed. It is shown to boost the convergence of optimal paths learnt by the agents as compared to traditional QLearning. Different foraging tasks are solved requiring varying degree of collective and individual efforts using the new proposed mechanism. The model enables us to characterize the ability of agents to solve complex foraging tasks rapidly and effectively.