The DOI matches its Crossref record; 32 of the 34 references it cites were found in a registry or an index.
End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Tasks
Open accessAbstract
Reinforcement Learning (RL) algorithms have found limited success beyond simulated applications, and one main reason is the absence of safety guarantees during the learning process. Real world systems would realistically fail or break before an optimal controller can be learned. To address this issue, we propose a controller architecture that combines (1) a model-free RL-based controller with (2) model-based controllers utilizing control barrier functions (CBFs) and (3) online learning of the unknown system dynamics, in order to ensure safety during learning. Our general framework leverages the success of RL algorithms to learn high-performance controllers, while the CBF-based controllers both guarantee safety and guide the learning process by constraining the set of explorable polices. We utilize Gaussian Processes (GPs) to model the system dynamics and its uncertainties. Our novel controller synthesis algorithm, RL-CBF, guarantees safety with high probability during the learning process, regardless of the RL algorithm used, and demonstrates greater policy exploration efficiency. We test our algorithm on (1) control of an inverted pendulum and (2) autonomous carfollowing with wireless vehicle-to-vehicle communication, and show that our algorithm attains much greater sample efficiency in learning than other state-of-the-art algorithms and maintains safety during the entire learning process.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Reinforcement learning (RL) algorithms have been increasingly adopted by the robotics community over the past years to control complex robots or multi-robot systems [1], [2], or provide end-to-end policies from perception to control [3].”
From Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey (2020) · cites this paper as [3]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Gaussian Processes for Machine Learning
Rasmussen, C. E., & Williams, C. K. I. (2005) · doi:10.7551/mitpress/3206.001.0001 · Cited by 11,705
Continuous control with deep reinforcement learning
Lillicrap, T. P., Hunt, J. J., Pritzel, A., Heess, N., Erez, T., Tassa, Y., Silver, D., & Wierstra, D. (2015) · arXiv · doi:10.48550/arxiv.1509.02971 · Cited by 10,782
(1995) · Cited by 10,558
Trust region policy optimization
Schulman, J., Levine, S., Moritz, P., Jordan, M. I., & Abbeel, P. (2015) · arXiv · doi:10.48550/arxiv.1502.05477 · Cited by 4,578 · cited 13×
“Furthermore, if we use TRPO for the RL algorithm, then the control law Proof.”
MethodInfluentialDeterministic Policy Gradient Algorithms
Silver, D., Lever, G., Heess, N., Degris, T., Wierstra, D., & Riedmiller, M. (2014) · HAL · Cited by 3,275
Control Barrier Function Based Quadratic Programs for Safety Critical Systems
Ames, A. D., Xu, X., Grizzle, J. W., & Paulo Tabuada, B. (2016) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2016.2638961 · Cited by 2,920 · cited 1×
“The existence of a CBF implies that there exists a deterministic controller u CBF : S → A such that the set C is forward invariant for system (1) (Agrawal and Sreenath 2017; Ames et al. 2017).”
BackgroundApprenticeship learning via inverse reinforcement learning
Abbeel, P., & Ng, A. Y. (2004) · International Conference on Machine Learning · doi:10.1145/1015330.1015430 · Cited by 2,855 · cited 1×
“Previous approaches to safe reinforcement learning include reward-shaping , policy optimization with constraints (Gaskett 2003; Moldovan and Abbeel 2012; Achiam et al. 2017; Wachi et al. 2018), or teacher advice (Abbeel and Ng 2004; Abbeel, Coates, and Ng 2010; Tang et al. 2010).”
Background(2015) · Journal of Machine Learning Research · Cited by 1,119
Benchmarking Deep Reinforcement Learning for Continuous Control
Duan, Y., Chen, X., Houthooft, R., Schulman, J., & Abbeel, P. (2016) · arXiv · doi:10.48550/arxiv.1604.06778 · Cited by 1,114 · cited 1×
“This method has been successfully applied to continuous control tasks (Duan et al. 2016; Lillicrap et al. 2015) where controllers have learned to stabilize complex robots (after many policy iterations).”
MethodReinforcement learning of motor skills with policy gradients
Peters, J., & Schaal, S. (2008) · Neural Networks · doi:10.1016/j.neunet.2008.02.003 · Cited by 831
Cited by
Sorted by citations
Items 501–600 of 603
Németh, B., & Gáspár, P. (2023) · Advances in industrial control · doi:10.1007/978-3-031-30537-5_1
SSRL: A Safe and Smooth Reinforcement Learning Approach for Collision Avoidance in Navigation
Zhang, R., Yang, J., Liang, Y., Lu, S., & Zhang, L. (2023) · Conference on Fully Actuated System Theory and Applications (FASTA) · doi:10.1109/cfasta57821.2023.10243245
Robust and Cheap Safety Measure for Exoskeletal Learning Control with Estimated Uniform PAC (EUPAC)
Weiske, F., & Jäkel, J. (2023) · European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN) · doi:10.14428/esann/2023.es2023-40
Efficient Re-synthesis of Control Barrier Function via Safe Exploration
Zehfroosh, A., Kong, Z., & Vougioukas, S. G. (2023) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc49753.2023.10384242
“CBF s have wide range of applications in safety-critical systems [2] among which one of the interesting ones is serving as a safety layer coupled with a learning-based control process so that the safety is guaranteed during the learning stage [3]–[5].”
BackgroundGestión segura de la batería de EVs utilizando aprendizaje por refuerzo
Trimboli, M., Antonelli, N., Avila, L., & Paula, M. D. (2024) · Programación matemática y software · doi:10.30973/progmat/2024.16.1/4
“Una forma de abordar el problema del SRL es guiar el proceso de aprendizaje restringiendo el conjunto de políticas explorables [11].”
BackgroundJin, C., Ma, X., Ren, T., Lin, W., & Ding, Z. (2024) · ACM International Conference on Hybrid Systems: Computation and Control (HSCC) · doi:10.1145/3641513.3650137
S2RU: scene semantic reconstruction for unmanned ground vehicle virtual-real integration
Ma, B., Wang, X., Xie, S., & Xiao, W. (2024) · International Conference on Electronic Information Engineering and Data Processing (EIEDP) · doi:10.1117/12.3032847
“DRL algorithms offer comprehensive strategies that encompass the entire process from perception to control, thereby equipping them with the ability to tackle highly complex problems 5 .”
Jana, R. K., Majumder, R., Bharathwaj, K. S., et al. (2024) · doi:10.1109/space63117.2024.10668292
Ensemble Probabilistic Model Predictive Safety Certification for Learning-Based Control
Güzelkaya, N., Gronauer, S., & Diepold, K. J. (2024) · IEEE Conference on Industrial Electronics and Applications (ICIEA) · doi:10.1109/iciea61579.2024.10664752
“In [9], it was shown that control barrier functions can be integrated into model-free RL to achieve safe exploration for continuous tasks. b) Learning the Uncertain Dynamics: This category focuses on enhancing the performance of learning-based controllers by effectively learning and adapting to…”
BackgroundDey, S., & Xu, H. (2024) · IEEE National Aerospace and Electronics Conference (NAECON) · doi:10.1109/naecon61878.2024.10670653
Approaching Single-Episode Survival Reinforcement Learning with Safety-Threshold Q-Learning
Perotto, F. S., Nargeot, M., & Ouahbi, A. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-77941-1_12
Gupta, P., & Jia, Y. (2023) · SAE technical papers on CD-ROM/SAE technical paper series · doi:10.4271/2024-01-4075
Review of Self-driving Car Based on NEAT Algorithm
Hotkar, O., Nambiar, P., Dhumane, A. V., Chiwhane, S., Sharma, A., Dharrao, D. S., & Meshram, V. A. (2024) · Lecture notes in networks and systems · doi:10.1007/978-981-97-6106-7_23
Data-Driven Safe Learning and Control with Safety Certificates
Ma, H., Yang, Y., Zou, W., et al. (2024) · Elsevier eBooks · doi:10.1016/b978-0-443-14081-5.00043-x
Impact of Communication on Project Success; with Mediating Role of Employee Efficiency
Gul, J., Rasheed, R., Usman, M., & Majid, A. (2025) · Review of Applied Management and Social Sciences · doi:10.47067/ramss.v8i1.431
Safe Switching Model-Free Value Iteration for General Nonlinear Systems
Lala, T. (2025) · IEEE Access · doi:10.1109/access.2025.3567496
“This stands as an improvement over safety checking mechanisms implemented in [17], [19], and [25], where this check is done online.”
Closed-Loop Safe Correction for Reinforcement Learning Policy
Yi, Z., Lv, Q., Chen, S., Liang, Y., & Dai, Y. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-4836-8_3
Risk‐aware safe reinforcement learning for control of stochastic linear systems
Esmaeili, B., Niknejad, N., & Modares, H. (2025) · Asian Journal of Control · doi:10.1002/asjc.3800
“…involves solving the following optimization problem where the constraints act as a shield, certifying the safety of the RL actions prior to deployment [36]. in which the constraint refers to a barrier certification constraint that ensures the set remains invariant, and u s signifies the…”
Zhang, H., Song, Y., Chen, G., et al. (2025) · Elsevier eBooks · doi:10.1016/b978-0-44-336492-1.00016-2
Zhang, H., Song, Y., Chen, G., & Yu, P. (2025) · Elsevier eBooks · doi:10.1016/b978-0-44-336492-1.00015-0
Liu, S., & Jingwei, B. (2025) · International Journal of Robust and Nonlinear Control · doi:10.1002/rnc.70049
“Recent works have integrated CBFs into various control strategies, but they often require initial admissible policies and strong assumptions [26], the extra large amount of computation [27], complicating implementation [28] and sometimes impairing optimality [29].”
Graceful Vehicle Collision Avoidance using a Second-Order Nonlinear Barrier Constraint
Moon, Y., Orosz, G., & Fathy, H. K. (2025) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc63710.2025.11107732
“…on the application of CBF methods to automotive safety includes collision and obstacle avoidance during racing [29], [30] and traffic congestion [31]; adaptive/connected cruise control with and without input delay [20], [32], [33]; human-vehicle trust dynamics [34]; lane-keeping control [4],…”
Self-citationStructure-Exploiting Reinforcement Learning for Networked Systems
Qu, G. (2025) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc63710.2025.11107771
Robust Perception-Based Navigation using PAC-NMPC with a Learned Value Function
Polevoy, A., Gonzales, M., Kobilarov, M., et al. (2025) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc63710.2025.11107521
“RL-CBF [40] learns a Gaussian Process model for the unknown dynamics to generate a CBF, which is then used as a safety filter on the RL-policy and to improve exploration efficiency.”
Resilient Safe Optimized Backstepping Control for High-Order Strict-Feedback System
Zhang, Y., Ji, R., & Ge, S. S. (2025) · doi:10.1109/icca65672.2025.11129817
“Safe reinforcement learning (SRL) thereafter is explored to provide suitable guarantees on the safety-related constraints during the control period, even during the learning period concerning the optimal learned policy [13].”
Leveraging Analytic Gradients in Provably Safe Reinforcement Learning
Walter, T., Markgraf, H., Külz, J., et al. (2025) · IEEE Open Journal of Control Systems · doi:10.1109/ojcsys.2025.3607845
“Withinthisfield of research, two common approaches for enforcing safety guarantees are control barrier functions [36], [37], [38], [39], [40] and reachability analysis [17], [41], [42].”
The Rule-Relocation Problem: From Referee Whistles to Thermostat Setpoints in Bio-Inspired AIPreprint
Arroyo, J. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5895763
Ergodicity and Control in Personalized Nutrition Via Constrained Learning DynamicsPreprint
Ghosh, R., Alahakoon, Y., Rathnayake, U., et al. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5367850
Pan, Q., Lian, J., Liu, X., & Wang, D. (2025) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3634606
“Theteacher-adviceframeworkiswidelyadoptedinRL[24], [25],[26].”
Kolhe, S. S., Patil, O. S., Dixon, W. E., et al. (2025) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc57313.2025.11312836
“The idea of using an estimate of the unknown dynamics to reduce conservativeness in a CBF-induced input constraint is used in [11] and [12], where the authors used Gaussian processes (GP) to estimate dynamics.”
LLM-in-the-Loop Adaptive Model Predictive Control for Sensorless PMSM Drives under UncertaintyPreprint
Wei, Y., & Zhang, Y. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6127029
Risk-Tunable Safe Adaptive Control for Nonlinear Systems Under Dynamical Uncertainties
Sharma, V., & Sivaranjani, S. (2026) · IEEE Open Journal of Control Systems · doi:10.1109/ojcsys.2026.3657701
“…condition, Adapted from [37]): The CBF h : X → R renders the dynamics (1) safe with respect to set S if there exist control inputs u t and a constant η ∈ [0 , 1] such that for all x t ∈ S , we have In Theorem 1, the constant η determines how conservatively the safety conditions are to be met [38].”
Yu, S., Li, S., Li, Y., et al. (2025) · CAA International Conference on Vehicular Control and Intelligence (CVCI) · doi:10.1109/cvci66304.2025.11348128
Zhang, S., Li, B., Chen, B., et al. (2026) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3644417
“Methods such as Control Barrier Function (CBF)- based quadratic programs (QP) [23] and Hamilton–Jacobi (HJ) reachability filters [24] act as correction layers after the RL controller, making minimal interventions to enforce safety constraints.”
DreamLander: A 6-DOF World Model Based Reinforcement Learning Planetary Lander
Nehma, G., Quinn, P. D., Tiwari, M., et al. (2026) · AIAA SciTech Forum · doi:10.2514/6.2026-1035
Fan, Z., Zhu, X., Tan, J., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6189653
Safe reinforcement learning for aerospace control via model-relaxed lyapunov stability
Chen, D., Yue, J., Zheng, T., et al. (2026) · Advances in Space Research · doi:10.1016/j.asr.2026.02.032
Alazzwi, A., Tshakwanda, P. M., Tsegaye, H. B., et al. (2026) · IEEE Annual Computing and Communication Workshop and Conference (CCWC) · doi:10.1109/ccwc67433.2026.11393718
From Safety to Guidance: CBF-Augmented Reinforcement Learning with Guiding Barrier Function
Nagata, K., & Arai, S. (2025) · IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) · doi:10.1109/iiai-aai-winter69777.2025.00031
Safe Reinforcement Learning Control for Maglev Train Levitation System with Robust Safety Assurance
Zhao, X., Sun, Y., Xu, J., et al. (2025) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc60802.2025.11423743
“[26] employed CBF to ensure safety during training, yet the modification of agent actions by CBF can destabilize the convergence of RL algorithms.”
Predictive Safeguarding and Performance Monitoring for Learning-Based Vehicle Motion Controllers
Ahmic, K., Ultsch, J., Brembeck, J., & Burschka, D. (2025) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc60802.2025.11423480
“In [10] and [11], control barrier functions (cf. [12]) have been applied in combination with RL to ensure safety by also adjusting the inputs such that the system is kept within a safe set of states.”
Hao, L., Liu, W., Huang, Q., et al. (2025) · IEEE PES Innovative Smart Grid Technologies - Asia (ISGT Asia) · doi:10.1109/isgtasia63446.2025.11431365
“To overcome this mismatch, Instead of using action projection or filtering as in [11], [19], we propose a safety value function (SVF) that transfers the CBF principle into a state-wise, value-based formulation.”
Sinaei, K., & Ebeigbe, D. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3678137 · Cited by 1
“…a CBF-based safety filter for RL agents that guides the policy towards the set of safe policies in addition to safeguarding the agent’s actions [15], Xiao et al. address the safety of learned models by introducing differentiable CBFs [16] and emphasize the need for safety guarantees when…”
Safe Event-Triggered Gaussian Process Learning for Barrier-Constrained Control
Lederer, A., Begzadić, A., Hirche, S., et al. (2026) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2026.3680122
“Notethatwefocusoncontinuous-timesystemsinthisarticleasCBF methodsforthediscrete-timesettingexhibitafundamentallydifferent structure[5].”
Safe Policy Optimization With Cost Practical Stability: A DQ-Learning Method
Wang, C., Liu, L., & Quan, Q. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3681154
“Additionally,wecompareagainstbarrier-basedmethods suchasPPO-CBFandTRPO-CBF[31],[43],whichintegrate controlbarrierfunctions(CBFs)assafetyfilterstoactively correctunsafeactions.”
Li, L. Y., wang, J. 王., Zhou, Z. J., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6544050
Tagliabue, E. (2026) · Springer tracts in advanced robotics · doi:10.1007/978-3-032-04247-7_6
Zakwan, M., & Giancarlo Ferrari-trecate, A. (2026) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2026.3683628
“Such limitations can be impractical, potentially leading to system failures during the training phase before an optimal policy can be found [27], [28].”
Isaac, E., George, A. J., Ioannou, I. I., et al. (2026) · Electronics · doi:10.3390/electronics15091945
“A novel discrete CBF-inspired filter [15,16] is introduced that augments the barrier-function safety test with a per episode visit memory to eliminate oscillation loops, and is paired with an empirical timeout/stagnation detector that converts blocked-goal, sealed-corridor and dead-end concave…”
InfluentialPeng, C., Guo, Y., & Ma, J. (2026) · International Journal of Robust and Nonlinear Control · doi:10.1002/rnc.70528
“Cheng et al. [32] developed a first model-free RL controller synthesis with CBF techniques to encode safe exploration.”
Ali, M., & Yousif, N. A. I. (2026) · doi:10.1109/icoecit68303.2026.11497523
Ding, J., Wu, T., Qian, Y., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-16855-9_10
A Transfer Learning Method for Safety Control of Linear Systems Under Input Saturation
Zhang, Q., Wu, C., & Wu, L. (2026) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2026.3692001
“Compared to existing methods that guide the training process in hazardous environments [4], [6], [7], [8], [9], this paradigm enhances safety more effectively by avoiding direct training in such environments.”
Guo, W., Li, D., Zhang, L., et al. (2026) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2026.3692963
“Cheng et al. [23] developed end-to-end safe reinforcement learning using CBF-based shields, while Taylor et al. [32] learned CBFs directly from data using support vector machines.”
Learning Safety-Guaranteed, Non-Greedy Control Barrier Functions Using Reinforcement Learning
Wijayatunga, M., Wallace, N., Sukkarieh, S., et al. (2026) · Proceedings - IEEE Aerospace Conference · doi:10.1109/aero66936.2026.11520115
“These methods either constrain the neural network’s actions using a pre-defined safety certificate [8], [9], learn safety certificates alongside a control policy [10], or derive certificates directly from observable quantities [11].”
Safe reinforcement learning for autonomous driving with behaviour correction mechanism
Zuo, Z., Liu, W., Wang, H., Li, Z., Li, P., & Wang, Y. (2026) · Journal of Control and Decision · doi:10.1080/23307706.2026.2679989
Bhagyalakshmi, L., Suman, S. K., Parvathy.K, et al. (2025) · doi:10.1109/ic-eeta66496.2025.11548192
Application of Risk-Driven Control in Transportation System
Li, K., Zhang, F., & (15202), L. W. (2026) · Lecture notes in electrical engineering · doi:10.1007/978-981-95-9350-7_7
Zhang, M., Zhang, H. M., Cao, Z., et al. (2026) · Conference on Fully Actuated System Theory and Applications (FASTA) · doi:10.1109/fasta70174.2026.11549200
Xu, Z., Yang, H., Liu, C., et al. (2026) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2026.3703136
“(1) Existing methods that enforce dynamically feasible velocity commands either (i) generate subgoals for a separate model-based controller to track [7], [8], [9] or (ii) constrain the action space to feasible actions [10], [11], [12], [13].”
Nemouchi, A., Bouzenada, A., & Saïdouni, D. E. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-19929-4_9
Autonomous Berthing in Mapless Environments via Soft- and Hard-Constrained Reinforcement Learning
Luo, J., Huang, H., Ding, Q., et al. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3706822
“It employs CBFs as a noninvasive module that corrects the output of an RL policy in real-time via optimization, typically quadratic programming (QP), to enforce safety constraints [27].”
Tan, J., Xue, S., Cao, H., et al. (2026) · doi:10.1109/icaisisas68969.2026.11567910
“Control barrier functions (CBFs) enforce state constraints through forward invariance of safe sets [11], and have been combined with RL via penalty-augmented rewards [12], constrained policy optimization [13], and safety filters [14].”
Cai, Y.-L., Zhang, X., Liu, S.-R., Lou, J., & Yuan, C.-H. (2026) · Nonlinear dynamics · doi:10.1007/s11071-026-12695-2
Ji, S., Ren, Y., Guo, Y., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2759-4_34
Ren, J., Chen, L., & Long, L. (2026) · Transactions of the Institute of Measurement and Control · doi:10.1177/01423312261465937
A Collaborative Safety Shield for Safe and Efficient CAV Lane Changes in Congested On-Ramp Merging
Hegde, B., & Bouroche, M. (2026) · IEEE Intelligent Vehicles Symposium · doi:10.1109/iv66570.2026.11623996
“The detailed derivation of this optimisation problem can be found in [13] and [24].”
Lyapunov-constrained reinforcement learning for provably stable robotic controlPreprint
Shinde, R., Shinde, K. D., Chaudhari, P., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7213796
Risk-Field Constrained Reinforcement Learning for Safe Autonomous Driving
Wang, X., & Zhang, Z. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31666-0_26
Risk-Aware Smooth Reinforcement Learning for Fixed-Wing UAV Aggressive Maneuvering
Zhu, H., Chen, M., Yan, C., et al. (2026) · International Conference on Control and Automation · doi:10.1109/icca69928.2026.11618194
“Finally, Neural CBF (NCBF) [14], [25] acts as a safety-critical baseline that enforces strict constraints.”
HYDRA-MC: Hybrid Dynamic Reconfigurable Architecture for Multi-Controller Autonomous Systems
Sriramulu, A., & Jayakanth, J. (2025) · doi:10.1109/dtda64656.2025.11622725
“The ensemble controller design mitigates single-point failure risk while enabling integration of classical and data-driven strategies [2], [8].”
InfluentialOnline MPC-Augmented Reinforcement Learning for Path Tracking Control of Autonomous Vehicles
Xu, Q., Cao, W., Wang, X., & Li, D. (2026) · International Conference on Control and Automation · doi:10.1109/icca69928.2026.11618160
A Hierarchical Reinforcement Learning Framework for Reach-Avoid Games with Obstacles
Yang, K., Chen, Y., & Chen, C. (2026) · International Conference on Control and Automation · doi:10.1109/icca69928.2026.11618257
“Utilizing constrained Markov decision processes (CMDPs), control barrier functions (CBF), and Lyapunov functions to reshape reward functions or project actions to ensure safety has become a mainstream approach [12]–[14].”
Tan, J., Wu, W., Chen, W., et al. (2026) · International Conference on Control and Automation · doi:10.1109/icca69928.2026.11618095
“Control Barrier Functions (CBFs) [18] provide a principled approach for enforcing forward invariance, and their combination with reinforcement learning has been explored in [19], [20].”
RiskBound: Risk-Aware Boundary-Guided Portfolio Optimization via Action Space Reshaping
Lee, H.-S., & Kang, U. (2026) · Knowledge Discovery and Data Mining · doi:10.1145/3770855.3817928
Peng, C., Qi, Q., Zhang, M., et al. (2026) · IEEE Systems Journal · doi:10.1109/jsyst.2026.3710033
“Recent control barrier function (CBF) technique and their variants offer formal assurances for safe control problems and have garnered significant attention within the realm of robotics research [21], [22], [23], [24], [25], [26], [27], [28].”
Applications of Reinforcement Learning for Autonomous Surgical Robotics: A Systematic Review
Shahid, M., Abdullah, A., Fatima, Z., Feroze, W., Torres-Ruiz, M. J., Saldaña-Pérez, M., Sanchéz-Mejorada, C. G., & Quintero, R. (2026) · Biomimetics · doi:10.3390/biomimetics11080577 · Cited by 1
Ahn, H.-J. (2026) · International Journal of Precision Engineering and Manufacturing-Green Technology · doi:10.1007/s40684-026-00905-6
Deep reinforcement learning-based thrust recovery control strategy for STOVL propulsion systems
Liu, S., Xu, F., Yang, S., et al. (2026) · Aerospace Science and Technology · doi:10.1016/j.ast.2026.113559
Safety-Enhanced COLREGs-Compliant Path Planning for USVs with a CBF-Based Safety Shield
Yun, S., Kim, H., Kwon, J., et al. (2026) · Journal of Marine Science and Engineering · doi:10.3390/jmse14161541
“In contrast, within the field of control theory, control filter frameworks utilizing Control Barrier Functions (CBFs) have been proposed to mathematically enforce forward-invariant safe regions within a system’s admissible control set [10].”
From stability to responsibility: the ethical evolution of AI-driven control systems
Khaki-Sedigh, A. (2026) · AI & Society · doi:10.1007/s00146-026-03302-4
Wei, W., & Liang, R. (2026) · Energy internet. · doi:10.1049/ein2.70035
Pan, G., Hou, B., & Chen, Y. (2026) · Drones · doi:10.3390/drones10100733
Safe Reinforcement Learning via Conservative Cost Shaping in Causal Latent RepresentationPreprint
Wang, H., & Jiang, L. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7355391
Sayghe, A., Sultan, A. B., & Mousa, M. A. (2026) · IEEE Access · doi:10.1109/access.2026.3728011 · Cited by 1
“Safe RL has produced multiple architectural patterns: Constrained Policy Optimization [11], Lyapunov-based safe RL [12], reward-constrained policy optimization [30], and CBF-shielded learning [13], [41].”
AI-driven control for dynamical systems: Taxonomy, guarantees, and applications
Liu, J., Ren, Y., Qian, S., Jin, X., & Li, F. (2026) · Discrete and Continuous Dynamical Systems - I Intelligence · doi:10.3934/dcdsi.r260301
Liu, H., Wang, R., Li, Y., et al. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3730349
“SRL-ORCA [16] uses Priority-ORCA as an external safety filter that checks the action proposed by the RL policy against an admissible velocity set and retains or modifies the action accordingly, while other works employ control barrier functions to constrain the policy search space [25].”
Li, D., & Shi, Y. (2026) · IEEE Access · doi:10.1109/access.2026.3729767
“CBF-based methods can enforce forward invariance when an appropriate safe set, dynamics model, and feasible online safety correction are available [11].”
Almamoori, H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7446074
Shielded reinforcement learning for industrial applications: A systematic literature survey
Odriozola-Olalde, H., Cano, F., Könighofer, B., Arana-Arexolaleiba, N., Zamalloa, M., & Perez-Cerrolaza, J. (2026) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2026.200729
Integrated design of safety and control systems for nonlinear safety-critical processes
Cui, X., Khodaverdian, A., Christofides, P. D., et al. (2026) · Process Safety and Environmental Protection · doi:10.1016/j.cherd.2026.09.021
Certified Safety Boundary Estimation for Autonomous Robotic Systems Under Uncertainty
Shi, H. (2026) · International Symposium on Control Engineering and Robotics (ISCER) · doi:10.1109/iscer70245.2026.11681164
“Safe reinforcement learning introduces CBF or CLF filters around a learned policy [8], [9].”
Learning safety-aware driving policies for unprotected left turns with barrier guidance
Ding, H., Bai, X., Fang, R., Bai, S., Cai, G., Zhuang, W., & Yin, G. (2026) · Mechatronics · doi:10.1016/j.mechatronics.2026.103618
Reactive and Safe Road User Simulations using Neural Barrier Certificates
Meng, Y., Qin, Z., & Fan, C. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636568
“Model-free safe RL methods [20], [21], [22], [23] can also be extended for multi-agent systems, but they draw less information from the expert trajectory.”
BackgroundSafe Reinforcement Learning Methods for Military Applications
Kim, S., Ahn, H., & Kim, J. (2026) · doi:10.1109/icmic70587.2026.11708604
“Depending on the application, safety may be formulated as a constraint on cumulative cost, a state-wise restriction, a risk-sensitive objective, a runtime action filter, a formally specified rule, explicit safety requirements, or deployment monitoring [10], [12], [13].”
Son, H., & Jung, S. (2026) · doi:10.1109/icmic70587.2026.11708708
“Control barrier function (CBF) safety filters [5]–[7] enforce safety by keeping the system inside a safe set, and have been scaled to large teams with graph neural networks [8]–[10] and combined with policy learning in safe RL [11]–[13].”
Shao, Z., Lei, L., Wang, P., et al. (2026) · Measurement and Control · doi:10.1177/00202940261494894
Dynamic safety-guaranteed control for direct current microgrids via safe reinforcement learning
Gong, Q., Zhou, L., Zhang, C., et al. (2026) · Transactions of the Institute of Measurement and Control · doi:10.1177/01423312261487888
Metadata
- Authors
- Cheng, R., Orosz, G., Murray, R. M., & Burdick, J. W.
- Affiliations
- California Institute of Technology (US)
- University of Michigan (US)
- Venue
- Proceedings of the AAAI Conference on Artificial Intelligence
- Published
- 2019-07-17
- Volume
- 33
- Issue
- 01
- Pages
- 3387-3395
- ISSN
- 2159-5399 (ISSN-L)
- 2374-3468
- OpenAlex
- W2966735560
- Semantic Scholar
- adcd4bfd88213da0c33d3cb7057411dd15b72c7a
- Cited by
- 611
- References
- 34
- Citations by year
- 2026: 79 · 2025: 84 · 2024: 79 · 2023: 102 · 2022: 84 · 2021: 112 · 2020: 60 · 2019: 10
- FWCI
- 25.44
- Type
- conference paper
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- www.aaai.org/ (unspecified)
- Locations
- Proceedings of the AAAI Conference on Artificial Intelligence (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- AAAI Publications (The Association for the Advancement of Artificial Intelligence (AAAI)) (submittedVersion)
- CaltechAUTHORS (California Institute of Technology) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Engineering › Control and Systems Engineering › Adaptive Dynamic Programming Control
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Formal Methods in Verification
- Keywords
- safe reinforcement learning, control barrier functions, safety-critical control, continuous control, model-based control, policy exploration, sample efficiency, inverted pendulum, vehicle-to-vehicle communication
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
