{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/arxiv-2604-19737","title":"Safe Continual Reinforcement Learning in Non-stationary Environments","arxiv_id":"2604.19737","date":"2026-04-21","proceeding":null,"authors":["Austin Coursey","Abel Diaz-Gonzalez","Marcos Quinones-Grueiro","Gautam Biswas"],"abstract":"Reinforcement learning (RL) offers a compelling data-driven paradigm for synthesizing controllers for complex systems when accurate physical models are unavailable; however, most existing control-oriented RL methods assume stationarity and, therefore, struggle in real-world non-stationary deployments where system dynamics and operating conditions can change unexpectedly. Moreover, RL controllers acting in physical environments must satisfy safety constraints throughout their learning and execution phases, rendering transient violations during adaptation unacceptable. Although continual RL and safe RL have each addressed non-stationarity and safety, respectively, their intersection remains comparatively unexplored, motivating the study of safe continual RL algorithms that can adapt over the system's lifetime while preserving safety. In this work, we systematically investigate safe continual reinforcement learning by introducing three benchmark environments that capture safety-critical continual adaptation and by evaluating representative approaches from safe RL, continual RL, and their combinations. Our empirical results reveal a fundamental tension between maintaining safety constraints and preventing catastrophic forgetting under non-stationary dynamics, with existing methods generally failing to achieve both objectives simultaneously. To address this shortcoming, we examine regularization-based strategies that partially mitigate this trade-off and characterize their benefits and limitations. Finally, we outline key open challenges and research directions toward developing safe, resilient learning-based controllers capable of sustained autonomous operation in changing environments.","url_abs":"https://arxiv.org/abs/2604.19737","url_pdf":"https://arxiv.org/pdf/2604.19737","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2604.19737","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2604.19737"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/MACS-Research-Lab/safe-crl","reach":null}],"summary":{"ran_draft_wrong":1,"ran_violates":1,"unverified":8},"by_repo_kind":{"found_in_text":{"samples":10,"ran":2,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":10,"samples":[{"code_sha256_prefix":"66df63cc965e3218","entry":"build_mlp_network","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/model.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/model.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"66df63cc965e3218"}},{"code_sha256_prefix":"9ca8b8e45f7fbebf","entry":"is_json_serializable","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/logger.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/logger.py","link_basis":"harvester_set","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"9ca8b8e45f7fbebf"}},{"code_sha256_prefix":"9edb7571305f7445","entry":"calculate_adv_and_value_targets","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/buffer.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/buffer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"9edb7571305f7445"}},{"code_sha256_prefix":"809acb410efac35d","entry":"colorize","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/logger.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/logger.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"809acb410efac35d"}},{"code_sha256_prefix":"2205dac1167fde2b","entry":"compute_fisher_info","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2205dac1167fde2b"}},{"code_sha256_prefix":"625c592552749a98","entry":"convert_json","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/logger.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/logger.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"625c592552749a98"}},{"code_sha256_prefix":"185c4bf8b0d68d6c","entry":"discount_cumsum","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/buffer.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/buffer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"185c4bf8b0d68d6c"}},{"code_sha256_prefix":"ab324c5144b1d1f7","entry":"get_hyperparameters","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"ab324c5144b1d1f7"}},{"code_sha256_prefix":"5b86ac2eed268258","entry":"make_sa_isaac_env","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/common/env.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/common/env.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"5b86ac2eed268258"}},{"code_sha256_prefix":"5fa56e888169271b","entry":"save_old_params","repo":"MACS-Research-Lab/safe-crl","repo_kind":"found_in_text","path":"code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","file_url":"https://github.com/MACS-Research-Lab/safe-crl/blob/HEAD/code/Safe-Policy-Optimization/safepo/single_agent/ppo_ewc_cost_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"5fa56e888169271b"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}