{"id":{"repo_id":"mit","oai_identifier":"oai:dspace.mit.edu:1721.1/150192"},"canonical_url":"https://search.dev.ndltd.org/etd/mit/oai:dspace.mit.edu:1721.1/150192","repository":{"repo_id":"mit","name":"MIT","base_url":"https://dspace.mit.edu/oai/request"},"display":{"title":"Discrete Approximate Information States in Partially Observable Environments","abstract":"The notion of approximate information states (AIS) was introduced in [31] as a methodology for learning task-relevant state representations for control in partially observable systems. They proposed particular learning objectives which attempt to reconstruct the cost and next state and provide a bound on the suboptimality of the closed-loop performance, but it is unclear whether these bounds are tight or actually lead to good performance in practice. Here we study this methodology by examining the special case of discrete approximate information states (DAIS). In this setting, we can solve for the globally optimal policy using value iteration for the DAIS model, allowing us to disambiguate the performance of the AIS objective from the policy search. Going further, for small problems with finite information states, we reformulate the DAIS learning problem as a novel mixed-integer program (MIP) and solve it to its global optimum; in the infinite information states case, we introduce clustering-based and end-to-end gradient-based optimization methods for minimizing the DAIS construction loss. We study DAIS in three partially observable environments and find that the AIS objective offers relatively loose bounds for guaranteeing monotonic performance improvement and is sufficient but not necessary for implementing optimal controllers. DAIS may even prove useful in practice by itself or as part of mixed discrete- and continuous-state representations, due to its ability to represent logical state, to its potential interpretabilty, and to the availability of these stronger algorithms.","abstract_html":"The notion of approximate information states (AIS) was introduced in [31] as a methodology for learning task-relevant state representations for control in partially observable systems. They proposed particular learning objectives which attempt to reconstruct the cost and next state and provide a bound on the suboptimality of the closed-loop performance, but it is unclear whether these bounds are tight or actually lead to good performance in practice. Here we study this methodology by examining the special case of discrete approximate information states (DAIS). In this setting, we can solve for the globally optimal policy using value iteration for the DAIS model, allowing us to disambiguate the performance of the AIS objective from the policy search. Going further, for small problems with finite information states, we reformulate the DAIS learning problem as a novel mixed-integer program (MIP) and solve it to its global optimum; in the infinite information states case, we introduce clustering-based and end-to-end gradient-based optimization methods for minimizing the DAIS construction loss. We study DAIS in three partially observable environments and find that the AIS objective offers relatively loose bounds for guaranteeing monotonic performance improvement and is sufficient but not necessary for implementing optimal controllers. DAIS may even prove useful in practice by itself or as part of mixed discrete- and continuous-state representations, due to its ability to represent logical state, to its potential interpretabilty, and to the availability of these stronger algorithms.","abstract_has_math":false,"creators":["Yang, Lujie"],"institution":"Massachusetts Institute of Technology","degree_name":"Master","degree_level":null,"degree_discipline":null,"degree_department":"Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science","school":null,"contributors":[],"advisors":["Tedrake, Russell L."],"committee_chairs":[],"committee_members":[],"year":2023,"date_issued":"2023-02","date_published":"2023-02","updated_at":"2026-07-22T22:21:26Z","subjects":[],"languages":[],"rights":["In Copyright - Educational Use Permitted","Copyright MIT"],"rights_urls":["http://rightsstatements.org/page/InC-EDU/1.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1721.1/150192","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Tedrake, Russell L."]},{"key":"dc:contributor.department","label":"Department","values":["Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science"]},{"key":"dc:creator","label":"Author","values":["Yang, Lujie"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2023-03-31T14:38:43Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2023-03-31T14:38:43Z"]},{"key":"dc:date.issued","label":"Date","values":["2023-02"]},{"key":"dc:publisher","label":"Institution","values":["Massachusetts Institute of Technology"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master","Master of Science in Electrical Engineering and Computer Science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright - Educational Use Permitted","Copyright MIT"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://rightsstatements.org/page/InC-EDU/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1721.1/150192"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The notion of approximate information states (AIS) was introduced in [31] as a methodology for learning task-relevant state representations for control in partially observable systems. They proposed particular learning objectives which attempt to reconstruct the cost and next state and provide a bound on the suboptimality of the closed-loop performance, but it is unclear whether these bounds are tight or actually lead to good performance in practice. Here we study this methodology by examining the special case of discrete approximate information states (DAIS). In this setting, we can solve for the globally optimal policy using value iteration for the DAIS model, allowing us to disambiguate the performance of the AIS objective from the policy search. Going further, for small problems with finite information states, we reformulate the DAIS learning problem as a novel mixed-integer program (MIP) and solve it to its global optimum; in the infinite information states case, we introduce clustering-based and end-to-end gradient-based optimization methods for minimizing the DAIS construction loss. We study DAIS in three partially observable environments and find that the AIS objective offers relatively loose bounds for guaranteeing monotonic performance improvement and is sufficient but not necessary for implementing optimal controllers. DAIS may even prove useful in practice by itself or as part of mixed discrete- and continuous-state representations, due to its ability to represent logical state, to its potential interpretabilty, and to the availability of these stronger algorithms."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["S.M."]},{"key":"dc:title","label":"Title","values":["Discrete Approximate Information States in Partially Observable Environments"]}]}],"canonical_facts":{"dc:contributor.advisor":["Tedrake, Russell L."],"dc:contributor.department":["Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science"],"dc:creator":["Yang, Lujie"],"dc:date.accessioned":["2023-03-31T14:38:43Z"],"dc:date.available":["2023-03-31T14:38:43Z"],"dc:date.issued":["2023-02"],"dc:description.abstract":["The notion of approximate information states (AIS) was introduced in [31] as a methodology for learning task-relevant state representations for control in partially observable systems. They proposed particular learning objectives which attempt to reconstruct the cost and next state and provide a bound on the suboptimality of the closed-loop performance, but it is unclear whether these bounds are tight or actually lead to good performance in practice. Here we study this methodology by examining the special case of discrete approximate information states (DAIS). In this setting, we can solve for the globally optimal policy using value iteration for the DAIS model, allowing us to disambiguate the performance of the AIS objective from the policy search. Going further, for small problems with finite information states, we reformulate the DAIS learning problem as a novel mixed-integer program (MIP) and solve it to its global optimum; in the infinite information states case, we introduce clustering-based and end-to-end gradient-based optimization methods for minimizing the DAIS construction loss. We study DAIS in three partially observable environments and find that the AIS objective offers relatively loose bounds for guaranteeing monotonic performance improvement and is sufficient but not necessary for implementing optimal controllers. DAIS may even prove useful in practice by itself or as part of mixed discrete- and continuous-state representations, due to its ability to represent logical state, to its potential interpretabilty, and to the availability of these stronger algorithms."],"dc:description.degree":["S.M."],"dc:identifier.uri":["https://hdl.handle.net/1721.1/150192"],"dc:publisher":["Massachusetts Institute of Technology"],"dc:rights":["In Copyright - Educational Use Permitted","Copyright MIT"],"dc:rights.uri":["http://rightsstatements.org/page/InC-EDU/1.0/"],"dc:title":["Discrete Approximate Information States in Partially Observable Environments"],"dc:type":["Thesis"],"thesis:degree_name":["Master","Master of Science in Electrical Engineering and Computer Science"]},"updated_at":"2026-07-22T22:21:26Z"}