{"title":"Plan-based reward shaping for reinforcement learning","authors":"M. Grzes, D. Kudenko","doi":"10.1109/IS.2008.4670492","DOIUrl":null,"url":null,"abstract":"Reinforcement learning, while being a highly popular learning technique for agents and multi-agent systems, has so far encountered difficulties when applying it to more complex domains due to scaling-up problems. This paper focuses on the use of domain knowledge to improve the convergence speed and optimality of various RL techniques. Specifically, we propose the use of high-level STRIPS operator knowledge in reward shaping to focus the search for the optimal policy. Empirical results show that the plan-based reward shaping approach outperforms other RL techniques, including alternative manual and MDP-based reward shaping when it is used in its basic form. We show that MDP-based reward shaping may fail and successful experiments with STRIPS-based shaping suggest modifications which can overcome encountered problems. The STRIPS-based method we propose allows expressing the same domain knowledge in a different way and the domain expert can choose whether to define an MDP or STRIPS planning task. We also evaluate the robustness of the proposed STRIPS-based technique to errors in the plan knowledge.","PeriodicalId":305750,"journal":{"name":"2008 4th International IEEE Conference Intelligent Systems","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2008-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"79","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2008 4th International IEEE Conference Intelligent Systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IS.2008.4670492","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 79
Abstract
Reinforcement learning, while being a highly popular learning technique for agents and multi-agent systems, has so far encountered difficulties when applying it to more complex domains due to scaling-up problems. This paper focuses on the use of domain knowledge to improve the convergence speed and optimality of various RL techniques. Specifically, we propose the use of high-level STRIPS operator knowledge in reward shaping to focus the search for the optimal policy. Empirical results show that the plan-based reward shaping approach outperforms other RL techniques, including alternative manual and MDP-based reward shaping when it is used in its basic form. We show that MDP-based reward shaping may fail and successful experiments with STRIPS-based shaping suggest modifications which can overcome encountered problems. The STRIPS-based method we propose allows expressing the same domain knowledge in a different way and the domain expert can choose whether to define an MDP or STRIPS planning task. We also evaluate the robustness of the proposed STRIPS-based technique to errors in the plan knowledge.