Using distributed w-learning for multi-policy optimization in decentralized autonomic systems

Distributed W-Learning (DWL) is a reinforcement learningbased algorithm for multi-policy optimization in agent-based systems. In this poster we propose the use of DWL for decentralized multi-policy optimization in autonomic systems. Using DWL, agents learn and exploit the dependencies between the policies that they are implementing, to collaboratively optimize the performance of an autonomic system. Our initial evaluation shows that DWL is a feasible algorithm for multi-policy optimization in decentralized autonomic systems. Our results show that a multi-policy collaborative DWL deployment outperforms individual single policy deployments, as well non-collaborative deployments. Categories and Subject Descriptors H.3.4 [Systems and Software]: Distributed systems; I.2.11 [Distributed Artificial Intelligence]:Multiagent systems General Terms Algorithms, Design, Experimentation Keywords Autonomic Computing, Reinforcement Learning, Decentralized Systems

• Type: Journal
• Year: 2009
• Where: ICAC
• Authors: Ivana Dusparic, Vinny Cahill
• Keywords: Applied Computing, Autonomic Systems, ICAC 2009, Multi-policy Optimization, Systems