A Hybrid Learning Strategy for Discovery of Policies of Action

This paper presents a novel hybrid learning method and performance evaluation methodology for adaptive autonomous agents. Measuring the performance of a learning agent is not a trivial task and generally requires long simulations as well as knowledge about the domain. A generic evaluation methodology has been developed to precisely evaluate the performance of policy estimation techniques. This methodology has been integrated into a hybrid learning algorithm which aim is to decrease the learning time and the amount of errors of an adaptive agent. The hybrid learning method namely Klearning, integrates the Q-learning and K Nearest-Neighbors algorithm. Experiments show that the K-learning algorithm surpasses the Q-learning algorithm in terms of convergence speed to a good policy.