Evaluating Q-Learning Policies for Multi-objective Foraging Task in a Multi-agent Environment

摘要

This paper evaluates the performances of the reported q-learning policies for multi-agent systems. A set of extensively used policies were identified in the open literature namely greedy, e-greedy, Boltzmann Distribution, Simulated Annealing and Probabiliy Matching. Five agents are modeled to search and retrieve pucks back to a home location in the environment under specified constraints. A number of simulation-based experiments was conducted and based on the numerical results that was obtained, the performances of the learning policies are discussed.