Learning and Adaption : A Comparison of Methods in Case of Navigation in an Artificial Robot World
Neural networks, reinforcement learning systems and evolutionary algorithms are widely used to solve problems in real-world robotics. We investigate learning and adaptation capabilities of agents and show that the learning time required in continual learning is shorter than that of learning from scratch under various learning conditions. We argue that agents using appropriate hybridization of learning and evolutionary algorithms show better learning and adaptation capability as compared to agents using learning algorithms only. We support our argument with experiments, where agents learn optimal policies in an artificial robot world
Use and reproduction:No Creative Commons license - The provisions of the German Copyright Act (UrhG) apply.Please note that individual components of the publication may be subject to other licensing or copyright conditions.