{"id":{"repo_id":"ohiolink","oai_identifier":"oai:etd.ohiolink.edu:case1354813154"},"canonical_url":"https://search.dev.ndltd.org/etd/ohiolink/oai:etd.ohiolink.edu:case1354813154","repository":{"repo_id":"ohiolink","name":"OhioLINK","base_url":"https://etd.ohiolink.edu/acprod/odb_etd/ws/oai/oai"},"display":{"title":"Integrating Deterministic Planning and Reinforcement Learning for Complex Sequential Decision Making","abstract":"This thesis presents a novel approach to solving decision-making problems in discrete, stochasticdomains. The method for solving these problems is often dictated by the availability of informationabout how the environment responds to actions taken by the agent. When the agentis given a model of the environment, it can plan out its actions beforehand, whereas an agentwithout a model must learn to differentiate good and bad decisions through direct experience.Until now, little attention has been paid to situations in which a model is only available for apart of the environment. We propose an algorithm which combines automated planning withhierarchical reinforcement learning in order to take advantage of the model when it is availableand sample from the environment when it is not. We prove that the same guarantees of optimalitythat apply to hierarchical reinforcement learning also apply to to this approach. Usingexperiments performed in two different domains, we demonstrate that hierarchically integratedplanning and reinforcement learning outperforms pure RL and pure planning hierarchies andthat this approach can scale to larger problems than are reasonably computable by other approaches.","abstract_html":"This thesis presents a novel approach to solving decision-making problems in discrete, stochasticdomains. The method for solving these problems is often dictated by the availability of informationabout how the environment responds to actions taken by the agent. When the agentis given a model of the environment, it can plan out its actions beforehand, whereas an agentwithout a model must learn to differentiate good and bad decisions through direct experience.Until now, little attention has been paid to situations in which a model is only available for apart of the environment. We propose an algorithm which combines automated planning withhierarchical reinforcement learning in order to take advantage of the model when it is availableand sample from the environment when it is not. We prove that the same guarantees of optimalitythat apply to hierarchical reinforcement learning also apply to to this approach. Usingexperiments performed in two different domains, we demonstrate that hierarchically integratedplanning and reinforcement learning outperforms pure RL and pure planning hierarchies andthat this approach can scale to larger problems than are reasonably computable by other approaches.","abstract_has_math":false,"creators":["Ernsberger, Timothy S."],"institution":"Case Western Reserve University School of Graduate Studies","degree_name":"Master of Sciences (Engineering)","degree_level":"masters","degree_discipline":"EECS - Computer and Information Sciences","degree_department":null,"school":null,"contributors":["Ray, Soumya"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2013,"date_issued":"2013-03-07","date_published":"2013-03-07","updated_at":"2026-07-24T03:35:52Z","subjects":["Artificial Intelligence","reinforcement learning","automated planning","Markov decision process"],"languages":["English"],"rights":["unrestricted","This thesis or dissertation is protected by copyright: some rights reserved. It is licensed for use under a Creative Commons license. Specific terms and permissions are available from this document's record in the OhioLINK ETD Center."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://rave.ohiolink.edu/etdc/view?acc_num=case1354813154","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Ray, Soumya"]},{"key":"dc:creator","label":"Author","values":["Ernsberger, Timothy S."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2013-03-07"]},{"key":"dc:publisher","label":"Institution","values":["Case Western Reserve University School of Graduate Studies / OhioLINK"]},{"key":"dc:type","label":"Dc Type","values":["Electronic Thesis or Dissertation"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["EECS - Computer and Information Sciences"]},{"key":"thesis:degree_level","label":"Degree Level","values":["masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Sciences (Engineering)"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Case Western Reserve University School of Graduate Studies"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Artificial Intelligence","reinforcement learning","automated planning","Markov decision process"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["English"]},{"key":"dc:rights","label":"Dc Rights","values":["unrestricted","This thesis or dissertation is protected by copyright: some rights reserved. It is licensed for use under a Creative Commons license. Specific terms and permissions are available from this document's record in the OhioLINK ETD Center."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://rave.ohiolink.edu/etdc/view?acc_num=case1354813154"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This thesis presents a novel approach to solving decision-making problems in discrete, stochasticdomains. The method for solving these problems is often dictated by the availability of informationabout how the environment responds to actions taken by the agent. When the agentis given a model of the environment, it can plan out its actions beforehand, whereas an agentwithout a model must learn to differentiate good and bad decisions through direct experience.Until now, little attention has been paid to situations in which a model is only available for apart of the environment. We propose an algorithm which combines automated planning withhierarchical reinforcement learning in order to take advantage of the model when it is availableand sample from the environment when it is not. We prove that the same guarantees of optimalitythat apply to hierarchical reinforcement learning also apply to to this approach. Usingexperiments performed in two different domains, we demonstrate that hierarchically integratedplanning and reinforcement learning outperforms pure RL and pure planning hierarchies andthat this approach can scale to larger problems than are reasonably computable by other approaches."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","600. KB"]},{"key":"dc:title","label":"Title","values":["Integrating Deterministic Planning and Reinforcement Learning for Complex Sequential Decision Making"]}]}],"canonical_facts":{"dc:contributor":["Ray, Soumya"],"dc:creator":["Ernsberger, Timothy S."],"dc:date":["2013-03-07"],"dc:description":["This thesis presents a novel approach to solving decision-making problems in discrete, stochasticdomains. The method for solving these problems is often dictated by the availability of informationabout how the environment responds to actions taken by the agent. When the agentis given a model of the environment, it can plan out its actions beforehand, whereas an agentwithout a model must learn to differentiate good and bad decisions through direct experience.Until now, little attention has been paid to situations in which a model is only available for apart of the environment. We propose an algorithm which combines automated planning withhierarchical reinforcement learning in order to take advantage of the model when it is availableand sample from the environment when it is not. We prove that the same guarantees of optimalitythat apply to hierarchical reinforcement learning also apply to to this approach. Usingexperiments performed in two different domains, we demonstrate that hierarchically integratedplanning and reinforcement learning outperforms pure RL and pure planning hierarchies andthat this approach can scale to larger problems than are reasonably computable by other approaches."],"dc:format":["application/pdf","600. KB"],"dc:identifier":["http://rave.ohiolink.edu/etdc/view?acc_num=case1354813154"],"dc:language":["English"],"dc:publisher":["Case Western Reserve University School of Graduate Studies / OhioLINK"],"dc:rights":["unrestricted","This thesis or dissertation is protected by copyright: some rights reserved. It is licensed for use under a Creative Commons license. Specific terms and permissions are available from this document's record in the OhioLINK ETD Center."],"dc:subject":["Artificial Intelligence","reinforcement learning","automated planning","Markov decision process"],"dc:title":["Integrating Deterministic Planning and Reinforcement Learning for Complex Sequential Decision Making"],"dc:type":["Electronic Thesis or Dissertation"],"thesis:degree_discipline":["EECS - Computer and Information Sciences"],"thesis:degree_level":["masters"],"thesis:degree_name":["Master of Sciences (Engineering)"],"thesis:institution_name":["Case Western Reserve University School of Graduate Studies"]},"updated_at":"2026-07-24T03:35:52Z"}