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 201–300 of 603
GoSafeOpt: Scalable safe exploration for global optimization of dynamical systems
Sukhija, B., Turchetta, M., Lindner, D., et al. (2023) · Artificial Intelligence · doi:10.1016/j.artint.2023.103922 · Cited by 12
Wang, C., Li, B., Song, L., et al. (2023) · Ocean Engineering · doi:10.1016/j.oceaneng.2023.115828 · Cited by 12
Reinforcement learning with model-based feedforward inputs for robotic table tennis
Ma, H., Büchler, D., Schölkopf, B., & Muehlebach, M. (2023) · Autonomous Robots · doi:10.1007/s10514-023-10140-6 · Cited by 17
“…augmenting the safety measures associated with real-world robot training, facilitating online training of robots in actual environments, even in the presence of mismatches between simulated and real-world systems (Garcia & Fernandez, 2015; Cheng et al., 2019; Ramya Ramakrishnan et al., 2020).”
BackgroundNeurosymbolic Motion and Task Planning for Linear Temporal Logic Tasks
Sun, X., & Shoukry, Y. (2024) · IEEE Transactions on Robotics · doi:10.1109/tro.2024.3392079 · Cited by 14
“Such a set of safe actions can be obtained through Hamilton–Jacobi analysis [35], [36] and barrier certificates [37], [38], [39], [40], [41], [42], [43].”
BackgroundTowards safe and sustainable reinforcement learning for real-time strategy games
Andersen, P., Goodwin, M., & Granmo, O. (2024) · Information Sciences · doi:10.1016/j.ins.2024.120980 · Cited by 12
Model-free Neural Lyapunov Control for Safe Robot Navigation
Xiong, Z., Eappen, J., Qureshi, A. H., & Jagannathan, S. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981632 · Cited by 14
“[29], [30], [31] can avoid all violations.”
BackgroundMethodInfluentialSafe Control for Nonlinear Systems Under Faults and Attacks via Control Barrier Functions
Zhang, H., Li, Z., & Clark, A. (2025) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2025.3571338 · Cited by 12
“CBFs for safe reinforcement learning were introduced in [43], [44], [45], [46], and [47].”
Safety Uncertainty in Control Barrier Functions using Gaussian Processes
Khan, M., Ibuki, T., & Chatterjee, A. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561504 · Cited by 12
“An RL algorithm combined with CBFs guarantees safety while learning the set of explorable policies as shown in [16].”
MethodCDT: Cascading Decision Trees for Explainable Reinforcement LearningPreprint
Ding, Z., Hernández-Leal, P., Ding, G. W., Li, C., & Huang, R. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.07553 · Cited by 19
“How to interpret the action choices in reinforcement learning (RL) policies remains a critical challenge, especially as the gradually increasing trend of applying RL in various domains involving transparency and safety (Cheng et al., 2019; Junges et al., 2016).”
BackgroundLi, Y., Das, S., Shamma, J. S., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.00411 · Cited by 12
“This paper is related with RL with constraints, which enjoys a lot of research but has limited results on both safety and non-asymptotic optimality guarantees (Marvi and Kiumarsi, 2021; Leurent et al., 2020; Fisac et al., 2018; Garcıa and Fernández, 2015; Cheng et al., 2019; Fulton and Platzer, 2018).”
BackgroundSadabadi, A. A., Rahimirad, Z., & Azimzadeh, H. (2022) · Journal of Environmental Planning and Management · doi:10.1080/09640568.2022.2043837 · Cited by 11
“behavior, generally called incentives and barriers (Alishahi, Moghaddam, and SheikhEl-Eslami 2012; Esmaieli and Ahmadian 2018; Cheng et al. 2019; Marvi and Kiumarsi 2021), must be identified.”
BackgroundMethodSafe Reinforcement Learning With Dead-Ends Avoidance and Recovery
Zhang, X., Zhang, H., Zhou, H., Huang, C. S., Zhang, D., Ye, C., & Zhao, J. (2023) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2023.3333248 · Cited by 11
“[11] [12] [13] assume a white-box or a black-box environment dynamics model is available, limiting policy optimizing in a safe policy set which is constructed based on the known model.”
BackgroundTowards Safe Load Balancing based on Control Barrier Functions and Deep Reinforcement Learning
Dinh, L. N., Quang, P. T. A., & Leguay, J. (2024) · IEEE/IFIP Network Operations and Management Symposium (NOMS) · doi:10.1109/noms59830.2024.10575399 · Cited by 10
“In this section, we propose a constrained policy optimization for load balancing based on Deep Reinforcement Learning (DRL) algorithms and a Control Barrier Function (CBF) [13].”
MethodSpecGuard: Specification Aware Recovery for Robotic Autonomous Vehicles from Physical Attacks
Dash, P., Chan, E., & Pattabiraman, K. (2024) · ACM SIGSAC Conference on Computer and Communications Security (CCS) · doi:10.1145/3658644.3690210 · Cited by 10
“While various safe Deep-RL approaches have been proposed [9, 15, 19], they are designed to handle environmental noise rather than attacks.”
Zhu, L., Tao, F., Fu, Z., et al. (2024) · Complex & Intelligent Systems · doi:10.1007/s40747-024-01698-4 · Cited by 11
“(19), the safe set C is forward invariant and asymptotically stable [24]. where γ is the attenuation coefficient from 0 to 1, which determines the repulsive strength of the safe set boundary on the system states (as the value of γ decreases, the repulsive strength of the safe set boundary to the…”
Zhang, Y., Tian, G., Wen, L., et al. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10802727 · Cited by 11
“This includes end-to-end safe reinforcement learning [9], end-to-end safe imitation learning [10], and integrating them into CLF-CBFs frameworks [11].”
Zhong, B., Lavaei, A., Cao, H., et al. (2021) · Nonlinear Analysis Hybrid Systems · doi:10.1016/j.nahs.2021.101110 · Cited by 11
Scukins, E., & Ögren, P. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561853 · Cited by 11
“Authors in [18] start with a CBF and use it to guarantee safety when applying RL.”
MethodSafe and Sample-Efficient Reinforcement Learning for Clustered Dynamic Environments
Chen, H., & Liu, C. (2021) · IEEE Control Systems Letters · doi:10.1109/lcsys.2021.3136486 · Cited by 11
“Nonetheless, the design rule can also be applied to CBF and we added a comparison of CBF+RL using the same φ.”
BackgroundMethodResultInfluentialSafe Learning Reference Governor: Theory and Application to Fuel Truck Rollover Avoidance
Liu, K., Li, N., Kolmanovsky, I. V., et al. (2021) · Journal of Autonomous Vehicles and Systems · doi:10.1115/1.4053244 · Cited by 11
“These approaches are typically either model-based [22] or require a robust controller for the nominal dynamics of the system [23].”
Medvedev, M., Pshikhopov, V., & Дмитриевич, Е. И. (2025) · Informatics and Automation · doi:10.15622/ia.24.3.1 · Cited by 10
Optimal battery charge with safe exploration
Trimboli, M., & Ávila, L. O. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.121697 · Cited by 10
Learning Volt-VAR Droop Curves to Optimally Coordinate Photovoltaic (PV) Smart Inverters
Glover, D., & Dubey, A. (2024) · IEEE Transactions on Industry Applications · doi:10.1109/tia.2024.3472655 · Cited by 10
“Promising work in [29], [30] discusses utilization of control barrier functions (CBFs) to improve safe learning under unknown system dynamics by establishing safe learning sets through boundary conditionals.”
Constrained Policy Optimization for Load Balancing
Kamri, A., Quang, P. T. A., Huin, N., & Leguay, J. (2021) · International Conference on the Design of Reliable Communication Networks (DRCN) · doi:10.1109/drcn51631.2021.9477375 · Cited by 9
“For instance, in future works, we may combine control barrier functions and DRL algorithms as proposed by Cheng et al. [26].”
MethodRMPs for Safe Impedance Control in Contact-Rich Manipulation
Shaw, S., Abbatematteo, B., & Konidaris, G. (2022) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra46639.2022.9811986 · Cited by 10
“Once the safety set is known, a method such as control barrier functions [17], [18] can be used to ensure that the policy stays within that set (termed forward set-invariance).”
MethodGunter, G., Nice, M. W., Bunting, M., et al. (2022) · doi:10.1109/di-cps56137.2022.00011 · Cited by 10
Model-free Learning for Safety-critical Control Systems: A Reference Governor Approach
Liu, K., Li, N., Kolmanovsky, I. V., et al. (2020) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc45564.2020.9147255 · Cited by 10
“These approaches are mostly model-based [3], [4], [5].”
MethodGaßmann, B., Pasch, F., Oboril, F., et al. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-55583-2_27 · Cited by 10
An Abstraction-based Method to Check Multi-Agent Deep Reinforcement-Learning Behaviors
Mqirmi, P. E., Belardinelli, F., & León, B. G. (2021) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/sral5637 · Cited by 12
Proximal policy optimization for formation navigation and obstacle avoidance
Sadhukhan, P., & Šelmić, R. R. (2022) · International Journal of Intelligent Robotics and Applications · doi:10.1007/s41315-022-00245-z · Cited by 10
“It is possible to completely avoid collisions and maintain an absolute safe distance in between agents and obstacles through the use of control barrier functions while training the policy, as done in Cheng et al. (2019)).”
BackgroundCorrect-by-Construction Runtime Enforcement in AI – A Survey
Könighofer, B., Bloem, R., Ehlers, R., & Pek, C. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-22337-2_31 · Cited by 18
Reactive and Safe Road User Simulations using Neural Barrier CertificatesPreprint
Meng, Y., Qin, Z., & Fan, C. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.06689 · Cited by 10
“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.”
BackgroundLearning Adaptive Safety for Multi-Agent Systems
Berducci, L., Yang, S., Mangharam, R., & Grosu, R. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10611037 · Cited by 11
“Problem (5) is tipically non-convex, and prior work focused on linear CBFs and convex formulations, for improved efficiency and optimal solutions [4].”
BackgroundA Framework for Online Updates to Safe Sets for Uncertain Dynamics
Shih, J. C., Meier, F., & Rai, A. (2020) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros45743.2020.9341606 · Cited by 9
“Constrained optimization, for example with barrier functions, can be used to certify probabilistic safety [6] or guide exploration to safe regions [7].”
BackgroundLearning Collision-free and Torque-limited Robot Trajectories based on Alternative Safe Behaviors
Kiemel, J. C., & Kröger, T. (2022) · IEEE-RAS International Conference on Humanoid Robots (Humanoids) · doi:10.1109/humanoids53995.2022.10000077 · Cited by 8
“Such a situation can be avoided by using control barrier functions (CBFs) [10]–[12], controlled invariant sets (CISs) [13], [14] or by ensuring the existence of a safe backup trajectory at each decision step [15], [16].”
BackgroundAsarkaya, A. S., Aksaray, D., & Yazıcıoğlu, Y. (2021) · International Conference on Unmanned Aircraft Systems (ICUAS) · doi:10.1109/icuas51884.2021.9476694 · Cited by 9
“Some works (e.g., [33]) have introduced barrier functions to avoid unsafe actions during the learning; however, they cannot accommodate complex constraints (e.g., temporal logics).”
BackgroundModel-based Reinforcement Learning with Provable Safety Guarantees via Control Barrier Functions
Zhang, H., Li, Z., & Clark, A. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561253 · Cited by 9
“In [9], a CBF construction for model-free RL was proposed that leverages bounds on the error of Gaussian processes (GPs) to ensure safe exploration.”
BackgroundMethodResultInfluentialDynamic Shielding for Reinforcement Learning in Black-Box Environments
Waga, M., Castellano, E., Pruekprasert, S., et al. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19992-9_2 · Cited by 9
Xiao, W., He, T., Dolan, J. M., & Shi, G. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10611340 · Cited by 13
“We adopt affine barrier functions as [21]: h = px + q, (p ∈ R, q ∈ R), for the simplicity of computation.”
BackgroundMethodData-Driven Estimation of Infinitesimal Generators of Stochastic Systems
Nejati, A., Lavaei, A., Soudjani, S., et al. (2021) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2021.08.511 · Cited by 9
Decentralized Safe Multi-agent Stochastic Optimal Control using Deep FBSDEs and ADMM
Pereira, M. A., Saravanos, A. D., So, O., et al. (2022) · doi:10.15607/rss.2022.xviii.055 · Cited by 9
“Therefore, very recently, the focus has shifted to the area of safe reinforcement learning [46, 12], where safety is required during the entire training process.”
BackgroundQuantifying Safety of Learning-based Self-Driving Control Using Almost-Barrier Functions
Qin, Z., Weng, T.-W., & Gao, S. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9982058 · Cited by 9
“The work in [24] uses control barrier functions to ensure safety of learned”
BackgroundDong, Y., Zhao, X., & Huang, X. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981794 · Cited by 9
“Since first proposed in 2013 [3], Deep Reinforcement Learning (DRL) has received significant attention in many applications [4]–[6].”
BackgroundContingency-constrained economic dispatch with safe reinforcement learning
Eichelbeck, M., Markgraf, H., & Althoff, M. (2022) · International Conference on Machine Learning and Applications · doi:10.1109/icmla55696.2022.00103 · Cited by 10
“where C(st, at) is the system cost from (3), α, β ∈ [0, 1] are weighting coefficients, and r t = ‖at − at ‖ is a safety correction penalty depending on the safe action at returned by the safety layer [15].”
BackgroundOnline Learning of non-Markovian Reward Models
Rens, G., Raskin, J.-F., Reynouard, R., et al. (2021) · International Conference on Agents and Artificial Intelligence (ICAART) · doi:10.5220/0010212000740086 · Cited by 9
“Some work has been done on safe learning/exploration (Turchetta et al., 2016; Hasanbeig et al., 2019b; Cheng et al., 2019) We expect that more complex non-Markovian reward environments will require intelligent exploration strategies.”
BackgroundZheng, Y., Yan, R., Jia, B., et al. (2024) · Physica A Statistical Mechanics and its Applications · doi:10.1016/j.physa.2024.130137 · Cited by 8
Safe Reinforcement Learning of Robot Trajectories in the Presence of Moving Obstacles
Kiemel, J. C., Righetti, L., Kröger, T., et al. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3488402 · Cited by 8
“More generally, techniques from model-based control can be used to increase safety when using RL [13], [14], [15].”
MethodBi, C., Liu, D., Zhang, X., Zhu, S., Li, S., Fu, C., & Lü, C. (2025) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2025.3587536 · Cited by 8
“Despite these advantages, constructing CBFs in practical systems remains challenging, as barrier function-based approaches either require strong assumptions or simplifications [26] or depend on accurate system modeling.”
Enforcing Cooperative Safety for Reinforcement Learning-Based Mixed-Autonomy Platoon Control
Zhou, J., Yan, L., Liang, J., et al. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3627592 · Cited by 8
“The main safe RL approaches include Constrained Markov Decision Process (CMDP) approaches [31], reachability analysis-based methods [32], and CBF-based approaches [33], [34], [35], [36], [37].”
A Survey on Learning Motion Planning and Control for Mobile Robots: Toward Embodied Intelligence
Wang, M., Niu, Y., Wang, B., Zhang, W., & Wang, C. (2026) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2026.3656889 · Cited by 9
“11(b)], typically implemented using control barrier functions (CBFs) [165] or MPC [32].”
Model-Free Reinforcement Learning for Stochastic Games with Linear Temporal Logic Objectives
Bozkurt, A. K., Wang, Y., Zavlanos, M. M., & Pajić, M. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561989 · Cited by 15
“Yet, despite these advances, there is a need to provide RL methods with robustness and safety guarantees to allow their use in real-world systems, particularly in cyber-physical [4], [5], [6], [7], [8] and robotic systems [9], [10], [11].”
BackgroundChriat, A. E., & Sun, C. (2023) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2023.3322088 · Cited by 8
“[2], [3], [4], [5], [6], [7], [8], [9] attracts much attention for safety assurance and explorations by using CBF as the safety shield.”
BackgroundEnhancing Safety in Learning from Demonstration Algorithms via Control Barrier Function Shielding
Yang, Y. F., Chen, L., Zaidi, Z., et al. (2024) · ACM/IEEE International Conference on Human-Robot Interaction (HRI) · doi:10.1145/3610977.3635002 · Cited by 8
“In the context of RL safety, researchers have investigated the utilization of expert-designed CBFs to synthesize control policies that confine the system within safe states [15, 16, 29, 30, 35].”
BackgroundEnforcing Constraints over Learned Policies via Nonlinear MPC: Application to the Pendubot
Turrisi, G., Carlos, B. B., Cefalo, M., Modugno, V., Lanari, L., & Oriolo, G. (2020) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2020.12.2426 · Cited by 10
A Joint Imitation-Reinforcement Learning Framework for Reduced Baseline Regret
Dey, S., Pendurkar, S., Sharon, G., & Hanna, J. P. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636294 · Cited by 8
“Methods relying on a model assume that the model is available beforehand [20], [21] while others learn the model online [15], [17], [22], [23].”
Run Time Assured Reinforcement Learning for Safe Satellite Docking
Dunlap, K. M., Mote, M., Delsing, K., & Hobbs, K. L. (2022) · AIAA SciTech Forum · doi:10.2514/6.2022-1853 · Cited by 8
Meng, X., & Chan, A. H. S. (2022) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph191912799 · Cited by 8
“Therefore, using graphical presentation and example illustration is recommended in areas that are difficult to understand while conducting on-job campaigns and safety courses, preferably with rewards and incentives for workers to ensure the learning initiative, such as safety performance reward or paid learning time […”
MethodSample-Efficient Safe Learning for Online Nonlinear Control with Control Barrier Functions
Luo, W., Sun, W. Y., & Kapoor, A. (2022) · Springer proceedings in advanced robotics · doi:10.1007/978-3-031-21090-7_25 · Cited by 8
Safety Filtering for Reinforcement Learning-based Adaptive Cruise Control
Hailemichael, H., Ayalew, B., Kerbel, L., Ivančo, A., & Loiselle, K. (2022) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2022.10.276 · Cited by 10
“The intuitive one is to come up with a handcrafted CBF considering the dynamics of the system and the action bounds associated with it (Xu et al., 2018; Ames et al., 2014; Cheng et al., 2019).”
BackgroundInstructed Reinforcement Learning Control of Safe Autonomous J-Turn Vehicle Maneuvers
Arab, A., & Yi, J. (2021) · IEEE/ASME International Conference on Advanced Intelligent Mechatronics · doi:10.1109/aim46487.2021.9517458 · Cited by 8
“Similarly, control barrier functions are used to guarantee safety with high probability during the learning process [11].”
BackgroundMethodSafe Interactive Model-Based LearningPreprint
Gallieri, M., Salehian, S. S. M., Toklu, N. E., et al. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1911.06556 · Cited by 8
CROP: Certifying Robust Policies for Reinforcement Learning through Functional SmoothingPreprint
Wu, F., Li, L., Huang, Z., Vorobeychik, Y., Zhao, D. X., & Li, B. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.09292 · Cited by 8
“As more and more safety-critical applications are being deployed in real-world [13, 22, 11, 1], how to test and improve their robustness before massive production is of great importance.”
BackgroundA Survey of Demonstration LearningPreprint
Correia, A. R. D. S. P., & Alexandre, L. A. (2023) · Robotics Auton. Syst. · doi:10.2139/ssrn.4390650 · Cited by 56
“In RL, existing algorithms tackle safety by specifying constraints that the agent can’t violate during execution [25, 136, 24, 112].”
BackgroundLearning Agile Quadrotor Flight in Restricted Environments With Safety Guarantees
Chen, S., Li, Y., Lou, Y., et al. (2024) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2024.3355688 · Cited by 7
Safe Autonomous Driving with Latent Dynamics and State-Wise Constraints
Wang, C., & Wang, Y. (2024) · Sensors · doi:10.3390/s24103139 · Cited by 7
Benyamen, H., Chowdhury, M., & Keshmiri, S. (2024) · Journal of Dynamic Systems Measurement and Control · doi:10.1115/1.4065804 · Cited by 8
An Admittance Parameter Optimization Method Based on Reinforcement Learning for Robot Force Control
Hu, X., Liu, G., Ren, P., et al. (2024) · Actuators · doi:10.3390/act13090354 · Cited by 7
Collision-Free Path Planning for Multiple Drones Based on Safe Reinforcement Learning
Chen, H., Huang, D., Wang, C., et al. (2024) · Drones · doi:10.3390/drones8090481 · Cited by 7
Differential High Order Control Barrier Function-Based Safe Reinforcement Learning
Kong, X., Xia, Y., Sun, Z., et al. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3575310 · Cited by 7
“To ensure safety, CBF-based safe RL methods have been studied in [10], [11], ensuring the RL agent satis-fies safety constraints during training and deployment.”
InfluentialLong- and Short-Term Constraint-Driven Safe Reinforcement Learning for Autonomous Driving
Hu, X., Chen, P., Wen, Y., Tang, B., & Chen, L. F. (2026) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2025.3648508 · Cited by 19
“Many safe RL methods [15], [16], [17], [18], [19], [20] use the expected cost, similar to the expected reward, as a secondary objective that mainly includes descriptions of decision risks or losses, which is a long-term objective constraint.”
Data‐driven safe gain‐scheduling control
Modares, A., Sadati, N., & Modares, H. (2023) · Asian Journal of Control · doi:10.1002/asjc.3169 · Cited by 7
Safe Control Using High-Order Measurement Robust Control Barrier Functions
Oruganti, P. S., Naghizadeh, P., & Ahmed, Q. (2023) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc55779.2023.10155975 · Cited by 7
“They have been successfully implemented in several applications including bipedal robots control [5], [6], automotive safety [1], and machine learning applications [7]–[9].”
BackgroundXie, J., Zhao, S., Hu, L., & Gao, H. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127225 · Cited by 7
“Some methods impose CBFs as constraints on the learnt policy via a non-invasive approach [16], [17], [18].”
Structured Learning of Safety Guarantees for the Control of Uncertain Dynamical Systems
Beaudoin, M.-A., & Boulet, B. (2022) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2022.3148212 · Cited by 7
“Several studies use a Gaussian process (GP) [30], [31] as a nonparametric model to both learn (or correct) a nominal model of the dynamics as well as bound its uncertainty [18], [19], [32], [33], [34].”
BackgroundLyapunov-based economic model predictive control for online model discrimination
Oyama, H., & Durand, H. (2022) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2022.107769 · Cited by 7
Safe reinforcement learning for dynamical systems using barrier certificates
Zhao, Q., Zhang, Y., & Li, X. (2022) · Connection Science · doi:10.1080/09540091.2022.2151567 · Cited by 7
“Cheng et al. (2019) propose an end-to-end architecture for safe reinforcement learning that combines the model-free reinforcement learning-based controller, model-based controller and online learningofunknownsystemdynamics.”
BackgroundData-Driven Stability Verification of Homogeneous Nonlinear Systems with Unknown Dynamics
Lavaei, A., Esfahani, P. M., & Zamani, M. (2022) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc51059.2022.9992739 · Cited by 7
“, [1], [2], [3], [4], [5]), obtaining an accurate model for”
BackgroundSecurity-Aware Reinforcement Learning under Linear Temporal Logic Specifications
Cui, B., Zhu, K., Li, S., et al. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10160753 · Cited by 7
“Most of these algorithms modify policy before the policy is executed actually using physical [7], [8] or logical [9], [10] methods.”
BackgroundHamilton, N. P., Dunlap, K. M., Johnson, T. T., & Hobbs, K. L. (2023) · IEEE International Conference on Space Mission Challenges for Information Technology · doi:10.1109/smc-it56444.2023.00014 · Cited by 9
“The reward function was modified from the original used in [38] by adding a constant, 5, which was chosen in order to make a majority of the reward values positive.”
MethodBypassing the Simulation-to-Reality Gap: Online Reinforcement Learning Using a Supervisor
Evans, B., Betz, J., Zheng, H., Engelbrecht, H. A., Mangharam, R., & Jordaan, H. W. (2023) · International Conference on Advanced Robotics (ICAR) · doi:10.1109/icar58858.2023.10406465 · Cited by 7
“Control barrier functions and similar set theory techniques have been used in several safety-critical learning problems [22], but have been focused on applications where a safe setting can be assumed [23] or the dynamics can be simplified to linear (affine) equations [24].”
BackgroundRuntime-Safety-Guided Policy Repair
Zhou, W., Gao, R., Kim, B., Kang, E., & Li, W. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-60508-7_7 · Cited by 7
Formal safety verification of unknown continuous-time systems
Lavaei, A., Nejati, A., Jagtap, P., et al. (2021) · International Conference on Hybrid Systems: computation and control · doi:10.1145/3447928.3456661 · Cited by 7
“, [3, 4]), acquiring an accurate model for complex systems is always very challenging, time-consuming, and expensive.”
BackgroundAssured Deep Multi-Agent Reinforcement Learning for Safe Robotic Systems
Riley, J., Călinescu, R., Paterson, C., Kudenko⋆, D., & Banks, A. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-10161-8_8 · Cited by 7
Machine learning-based steering control for automated vehicles utilizing V2X communication
Avedisov, S. S., He, C. R., Takács, D., et al. (2021) · IEEE Conference on Control Technology and Applications (CCTA) · doi:10.1109/ccta48906.2021.9658972 · Cited by 7
“Rather than constructing separate decision making, planning, and control algorithms, here we take an “end-to-end” approach [2], [3].”
MethodSelf-citationBehavior Planning For Connected Autonomous Vehicles Using Feedback Deep Reinforcement Learning.Preprint
Han, S., & Miao, F. (2020) · arXiv (Cornell University) · Cited by 7
Safe Reinforcement Learning on the Constraint Manifold: Theory and Applications
Liu, P., Ammar, H. B., Peters, J., & Tateo, D. (2025) · IEEE Transactions on Robotics · doi:10.1109/tro.2025.3567477 · Cited by 6
“To address the first challenge, methods such as control barrier functions (CBFs) [10], [36], [37], [38] and reachability analysis [9], [39], [40], [41], [42] utilize prior .”
Zhu, K., Xue, T., & Zhang, T. (2025) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2025.3578326 · Cited by 10
“Insights from control theory have influenced state-wise safe RL, with some RL studies incorporating control barrier functions to ensure persistent safety and define feasible sets of states [43].”
Constraint-Guided Reinforcement Learning: Augmenting the Agent-Environment-Interaction
Spieker, H. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9533996 · Cited by 6
“Similar works have been based on using barrier functions [14] or provable formal theories of the acceptable behavior specifications of the RL agent [15], or focus on specifics of the RL environment, e.g. stochastic dynamics in robotics [16].”
Haider, T., Roscher, K., Herd, B., et al. (2024) · ACM/SIGAPP Symposium on Applied Computing (SAC) · doi:10.1145/3605098.3635931 · Cited by 6
“Another line of work aims to compensate for unsafe control actions via shielding or control-barrier functions [2, 16].”
BackgroundAutonomous Driving via Knowledge-Enhanced Safe Reinforcement Learning
Wang, C., Zhou, S., Wang, L., Lu, Z., Wu, C., Wen, X., & Shou, G. (2024) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2024.3412916 · Cited by 7
“These methods firstly identify the risky scenarios by the feasible region [21] and model-based constraints [22], [23], [24], then apply risk-averse policies, e.g., braking in front of the obstacles [21], [26], expert strategies [21], control barrier functions [22], [23], [24], and optimization [14].”
A Survey of Research on Stability Guarantee of Reinforcement Learning Automatic Control Problem
Gan, Q., Wu, F., & Zhao, J. (2021) · International Conference on Control, Robotics and Cybernetics (CRC) · doi:10.1109/crc52766.2021.9620159 · Cited by 6
Safe Distributional Reinforcement Learning
Zhang, J., & Weng, P. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-94662-3_8 · Cited by 6
“One main line of work [Turchetta et al., 2016, Berkenkamp et al., 2017, Wachi et al., 2018, Cheng et al., 2019] tries to prevent the choice of a bad action by learning a model.”
BackgroundSafe Hierarchical Navigation in Crowded Dynamic Uncertain Environments
Chen, H., Feng, S., Zhao, Y., et al. (2022) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc51059.2022.9992674 · Cited by 6
“Recent safe learning algorithms use CBF or SSA as action monitor to keep modifying actions generated from the policy and achieve a low collision rate for safety-critical tasks [13][35].”
MethodDetecting message spoofing attacks on smart vehicles
Ibrahim, M., & Safa, N. S. (2023) · Computer Fraud & Security · doi:10.12968/s1361-3723(23)70054-7 · Cited by 6
Alighanbari, S., & Azad, N. L. (2021) · IEEE Access · doi:10.1109/access.2021.3119915 · Cited by 7
Safe Reinforcement Learning for an Energy-Efficient Driver Assistance System
Hailemichael, H., Ayalew, B., Kerbel, L., et al. (2022) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2022.11.250 · Cited by 6
Connectivity Guaranteed Multi-robot Navigation via Deep Reinforcement Learning.
Lin, J., Yang, X., Zheng, P., & Cheng, H. (2019) · Conference on Robot Learning · Cited by 6
Safe Model-Based Reinforcement Learning Using Robust Control Barrier Functions.Preprint
Emam, Y., Glotfelter, P., Kira, Z., et al. (2021) · arXiv (Cornell University) · Cited by 6
Safe Autonomous Racing via Approximate Reachability on Ego-visionPreprint
Chen, B., Francis, J., Oh, J., Nyberg, E., & Herbert, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2110.07699 · Cited by 6
“While a number of works have similar dual-policy architecture [6, 13, 46], ours design is informed by HJ Reachability theory.”
BackgroundMethodInfluentialLiu, S., Yu, Z., & Gao, D. (2025) · International Journal of Systems Science · doi:10.1080/00207721.2025.2461016 · Cited by 5
“To validate our algorithm’s efficiency in terms of computational resources and time, we compare our method against a conventional QP-based RL method with CBFs (Cao et al., 2024; Cheng et al.,2019 ).”
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
