{"id":"https://openalex.org/W7138230318","doi":"https://doi.org/10.1109/tpami.2026.3674995","title":"Mirror Descent Safe Policy Optimization for Reinforcement Learning Agents","display_name":"Mirror Descent Safe Policy Optimization for Reinforcement Learning Agents","publication_year":2026,"publication_date":"2026-03-17","ids":{"openalex":"https://openalex.org/W7138230318","doi":"https://doi.org/10.1109/tpami.2026.3674995","pmid":"https://pubmed.ncbi.nlm.nih.gov/41843516"},"language":"en","primary_location":{"id":"doi:10.1109/tpami.2026.3674995","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tpami.2026.3674995","pdf_url":null,"source":{"id":"https://openalex.org/S199944782","display_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","issn_l":"0162-8828","issn":["0162-8828","1939-3539","2160-9292"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","pubmed"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5122119566","display_name":"Renzhi Lu","orcid":null},"institutions":[{"id":"https://openalex.org/I4210105785","display_name":"Intelligent Systems Research (United States)","ror":"https://ror.org/01reevc91","country_code":"US","type":"company","lineage":["https://openalex.org/I4210105785"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Renzhi Lu","raw_affiliation_strings":["School of Artificial Intelligence and Automation, Key Laboratory of Image Processing and Intelligent Control, Engineering Research Center of Autonomous Intelligent Unmanned Systems, Chinese Ministry of Education, Huazhong University of Science and Technology, Wuhan, China"],"raw_orcid":"https://orcid.org/0000-0001-6845-1711","affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence and Automation, Key Laboratory of Image Processing and Intelligent Control, Engineering Research Center of Autonomous Intelligent Unmanned Systems, Chinese Ministry of Education, Huazhong University of Science and Technology, Wuhan, China","institution_ids":["https://openalex.org/I4210105785"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059122235","display_name":"Ning Wu","orcid":"https://orcid.org/0000-0002-4951-6337"},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ning Wu","raw_affiliation_strings":["School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129748164","display_name":"Qingqing Xiong","orcid":null},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingqing Xiong","raw_affiliation_strings":["School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110604882","display_name":"Yifang Shi","orcid":null},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yifang Shi","raw_affiliation_strings":["School of Automation, Hangzhou Dianzi University, Hangzhou, China"],"raw_orcid":"https://orcid.org/0000-0003-1607-3629","affiliations":[{"raw_affiliation_string":"School of Automation, Hangzhou Dianzi University, Hangzhou, China","institution_ids":["https://openalex.org/I50760025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121979065","display_name":"Dongrui Wu","orcid":null},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dongrui Wu","raw_affiliation_strings":["Key Laboratory of the Ministry of Education for Image Processing and Intelligent Control, School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China"],"raw_orcid":"https://orcid.org/0000-0002-7153-9703","affiliations":[{"raw_affiliation_string":"Key Laboratory of the Ministry of Education for Image Processing and Intelligent Control, School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, China","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114763051","display_name":"Tao Yang","orcid":"https://orcid.org/0000-0003-0832-7457"},"institutions":[{"id":"https://openalex.org/I9224756","display_name":"Northeastern University","ror":"https://ror.org/03awzbc87","country_code":"CN","type":"education","lineage":["https://openalex.org/I9224756"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tao Yang","raw_affiliation_strings":["State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University, Shenyang, China"],"raw_orcid":"https://orcid.org/0000-0003-4090-8497","affiliations":[{"raw_affiliation_string":"State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University, Shenyang, China","institution_ids":["https://openalex.org/I9224756"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129731980","display_name":"Yaochu Jin","orcid":null},"institutions":[{"id":"https://openalex.org/I3133055985","display_name":"Westlake University","ror":"https://ror.org/05hfa4n20","country_code":"CN","type":"education","lineage":["https://openalex.org/I3133055985"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yaochu Jin","raw_affiliation_strings":["School of Engineering, Westlake University, Hangzhou, China"],"raw_orcid":"https://orcid.org/0000-0003-1100-0631","affiliations":[{"raw_affiliation_string":"School of Engineering, Westlake University, Hangzhou, China","institution_ids":["https://openalex.org/I3133055985"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5129657267","display_name":"Lihua Xie","orcid":null},"institutions":[{"id":"https://openalex.org/I172675005","display_name":"Nanyang Technological University","ror":"https://ror.org/02e7b5302","country_code":"SG","type":"education","lineage":["https://openalex.org/I172675005"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Lihua Xie","raw_affiliation_strings":["School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore"],"raw_orcid":"https://orcid.org/0000-0002-7137-4136","affiliations":[{"raw_affiliation_string":"School of Electrical and Electronics Engineering, Nanyang Technological University, Singapore","institution_ids":["https://openalex.org/I172675005"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.27363731,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"48","issue":"8","first_page":"9142","last_page":"9155"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.7462999820709229,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.7462999820709229,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12794","display_name":"Adaptive Dynamic Programming Control","score":0.12939999997615814,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.0215000007301569,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.621399998664856},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.42559999227523804},{"id":"https://openalex.org/keywords/descent","display_name":"Descent (aeronautics)","score":0.4009999930858612},{"id":"https://openalex.org/keywords/markov-decision-process","display_name":"Markov decision process","score":0.2924000024795532},{"id":"https://openalex.org/keywords/multi-agent-system","display_name":"Multi-agent system","score":0.28110000491142273},{"id":"https://openalex.org/keywords/minification","display_name":"Minification","score":0.27459999918937683}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6647999882698059},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.621399998664856},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6065999865531921},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.42559999227523804},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41100001335144043},{"id":"https://openalex.org/C2776637919","wikidata":"https://www.wikidata.org/wiki/Q624380","display_name":"Descent (aeronautics)","level":2,"score":0.4009999930858612},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3181000053882599},{"id":"https://openalex.org/C106189395","wikidata":"https://www.wikidata.org/wiki/Q176789","display_name":"Markov decision process","level":3,"score":0.2924000024795532},{"id":"https://openalex.org/C41550386","wikidata":"https://www.wikidata.org/wiki/Q529909","display_name":"Multi-agent system","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.27459999918937683},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.26460000872612},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.26019999384880066},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.2540000081062317},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2515999972820282}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1109/tpami.2026.3674995","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tpami.2026.3674995","pdf_url":null,"source":{"id":"https://openalex.org/S199944782","display_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","issn_l":"0162-8828","issn":["0162-8828","1939-3539","2160-9292"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","raw_type":"journal-article"},{"id":"pmid:41843516","is_oa":false,"landing_page_url":"https://pubmed.ncbi.nlm.nih.gov/41843516","pdf_url":null,"source":{"id":"https://openalex.org/S4306525036","display_name":"PubMed","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1299303238","host_organization_name":"National Institutes of Health","host_organization_lineage":["https://openalex.org/I1299303238"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE transactions on pattern analysis and machine intelligence","raw_type":null},{"id":"pmh:oai:dr.ntu.edu.sg:10356/211539","is_oa":false,"landing_page_url":"https://hdl.handle.net/10356/211539","pdf_url":null,"source":{"id":"https://openalex.org/S4306402609","display_name":"DR-NTU (Nanyang Technological University)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I172675005","host_organization_name":"Nanyang Technological University","host_organization_lineage":["https://openalex.org/I172675005"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Journal Article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/10","display_name":"Reduced inequalities","score":0.5321320295333862}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Embodied":[0],"intelligence":[1],"and":[2,57,102,155,179,194,223,259,287],"related":[3],"disciplines":[4],"have":[5],"identified":[6],"several":[7],"mechanisms":[8],"that":[9,98,250],"help":[10],"embodied":[11],"agents":[12],"learn":[13],"how":[14],"to":[15,126,140,187,196],"solve":[16],"complex":[17],"problems.":[18],"Reinforcement":[19],"learning":[20],"(RL)":[21],"is":[22,124,153,161,191],"one":[23],"of":[24,32,37,45,89,106,130,213,233,244],"the":[25,33,41,74,90,128,142,146,166,171,176,182,208,214,231,253,262,284],"most":[26],"promising":[27],"computational":[28],"approaches":[29],"toward":[30],"enhancement":[31],"learning-based":[34],"problem-solving":[35],"abilities":[36],"such":[38],"agents.":[39],"Given":[40],"recent":[42],"rapid":[43],"evolution":[44],"artificial":[46],"intelligence,":[47],"RL":[48,132],"has":[49],"become":[50],"a":[51,78,81,115,192,204,217,271],"keystone":[52],"technology,":[53],"accelerating":[54],"scientific":[55],"discoveries":[56],"also":[58],"finding":[59],"applications":[60],"in":[61,97],"many":[62],"other":[63,266],"domains.":[64],"In":[65,112,270],"RL,":[66],"an":[67,107,131,156,224,277],"agent":[68,108,289],"collects":[69],"data":[70],"when":[71],"interacting":[72],"with":[73,175],"environment,":[75],"which":[76,200],"optimizes":[77],"policy":[79,120,173,184,235],"ensuring":[80],"higher":[82],"return.":[83],"Further":[84],"improvement":[85,222],"requires":[86],"more":[87],"exploration":[88,105],"action":[91],"space.":[92,185],"However,":[93],"not":[94,202],"all":[95],"actions":[96],"space":[99,174],"are":[100],"safe":[101,119],"acceptable.":[103],"The":[104,134,237],"must":[109],"be":[110],"constrained.":[111],"this":[113],"work,":[114],"novel":[116,150],"mirror":[117,137],"descent":[118,138,164],"optimization":[121,139,151,159],"(MDSPO)":[122],"algorithm":[123,135],"proposed":[125],"ensure":[127],"safety":[129,147],"agent.":[133],"leverages":[136],"maximize":[141],"return":[143,221,255],"while":[144],"satisfying":[145],"constraint.":[148],"A":[149],"objective":[152],"formulated,":[154],"innovative":[157],"three-stage":[158],"strategy":[160],"employed-comprising":[162],"gradient":[163],"without":[165],"cost":[167,177,263],"constraint,":[168,178],"projection":[169,180],"onto":[170,181],"nonparametric":[172],"parametric":[183],"Compared":[186],"previous":[188],"methods,":[189],"MDSPO":[190,215,251,281],"simple":[193],"easy":[195],"implement":[197],"first-order":[198],"approach,":[199],"does":[201],"impose":[203],"hard":[205],"constraint":[206,228],"on":[207,220,227],"trust":[209],"region.":[210],"Theoretical":[211],"analysis":[212],"reveals":[216],"lower":[218],"bound":[219,226],"upper":[225],"violation":[229],"at":[230],"time":[232],"each":[234],"update.":[236],"numerical":[238],"results":[239],"obtained":[240],"from":[241],"two":[242],"sets":[243],"different":[245],"constrained":[246],"locomotive":[247],"experiments":[248],"demonstrate":[249],"improves":[252],"average":[254],"by":[256],"about":[257],"12%":[258],"better":[260],"satisfies":[261],"constraints":[264],"than":[265],"state-of-the-art":[267],"methods":[268],"do.":[269],"real-world":[272],"obstacle":[273],"avoidance":[274],"experiment":[275],"using":[276],"unmanned":[278],"surface":[279],"vessel,":[280],"both":[282],"finds":[283],"optimal":[285],"path":[286],"guarantees":[288],"safety.":[290]},"counts_by_year":[],"updated_date":"2026-07-07T06:14:10.401900","created_date":"2026-03-18T00:00:00"}