Vanya Markova
Ventseslav Shopov
Institute of Robotics, Bulgarian Academy of Science
Bulgaria
e-mail: markovavanya@yahoo.com
Abstract:
In this article, we discuss deep learning methods for reinforcement learning agents. We study the application of multi-agent deep reinforcement learning techniques in agents. The main hypothesis is that deep reinforcement learning and collective behaviour approach demonstrate better performance than classic reinforcement learning. So autonomous agents are capable of discovering good solutions to the problem at hand by cooperate with other learners.
Key words:
reinforcement learning
collective robots
knowledge transfer