Safe reinforcement learning (RL) aims to learn policies that satisfy certain constraints before deploying them to safety-critical applications. Previous primal-dual style approaches suffer from instability issues and lack optimality guarantees. This paper overcomes the issues from the perspective of probabilistic inference. We introduce a novel Expectation-Maximization approach to naturally incorporate constraints during the policy learning: 1) a provable optimal non-parametric variational distribution could be computed in closed form after a convex optimization (E-step); 2) the policy parameter is improved within the trust region based on the optimal variational distribution (M-step). The proposed algorithm decomposes the safe RL problem into a convex optimization phase and a supervised learning phase, which yields a more stable training performance. A wide range of experiments on continuous robotic tasks shows that the proposed method achieves significantly better constraint satisfaction performance and better sample efficiency than baselines. The code is available at https://github.com/liuzuxin/cvpo-safe-rl.
Reinforcement Learning (RL) agents in the real world must satisfy safety constraints in addition to maximizing a reward objective. Model-based RL algorithms hold promise for reducing unsafe real-world actions: they may synthesize policies that obey all constraints using simulated samples from a learned model. However, imperfect models can result in real-world constraint violations even for actions that are predicted to satisfy all constraints. We propose Conservative and Adaptive Penalty (CAP), a model-based safe RL framework that accounts for potential modeling errors by capturing model uncertainty and adaptively exploiting it to balance the reward and the cost objectives. First, CAP inflates predicted costs using an uncertainty-based penalty. Theoretically, we show that policies that satisfy this conservative cost constraint are guaranteed to also be feasible in the true environment. We further show that this guarantees the safety of all intermediate solutions during RL training. Further, CAP adaptively tunes this penalty during training using true cost feedback from the environment. We evaluate this conservative and adaptive penalty-based approach for model-based safe RL extensively on state and image-based environments. Our results demonstrate substantial gains in sample-efficiency while incurring fewer violations than prior safe RL algorithms. Code is available at: https://github.com/Redrew/CAP
Safe reinforcement learning is extremely challenging--not only must the agent explore an unknown environment, it must do so while ensuring no safety constraint violations. We formulate this safe reinforcement learning (RL) problem using the framework of a finite-horizon Constrained Markov Decision Process (CMDP) with an unknown transition probability function, where we model the safety requirements as constraints on the expected cumulative costs that must be satisfied during all episodes of learning. We propose a model-based safe RL algorithm that we call Doubly Optimistic and Pessimistic Exploration (DOPE), and show that it achieves an objective regret $\tilde{O}(|\mathcal{S}|\sqrt{|\mathcal{A}| K})$ without violating the safety constraints during learning, where $|\mathcal{S}|$ is the number of states, $|\mathcal{A}|$ is the number of actions, and $K$ is the number of learning episodes. Our key idea is to combine a reward bonus for exploration (optimism) with a conservative constraint (pessimism), in addition to the standard optimistic model-based exploration. DOPE is not only able to improve the objective regret bound, but also shows a significant empirical performance improvement as compared to earlier optimism-pessimism approaches.
Reinforcement Learning (RL) has been shown to be effective in many scenarios. However, it typically requires the exploration of a sufficiently large number of state-action pairs, some of which may be unsafe. Consequently, its application to safety-critical systems remains a challenge. An increasingly common approach to address safety involves the addition of a safety layer that projects the RL actions onto a safe set of actions. In turn, a difficulty for such frameworks is how to effectively couple RL with the safety layer to improve the learning performance. In this paper, we frame safety as a differentiable robust-control-barrier-function layer in a model-based RL framework. Moreover, we also propose an approach to modularly learn the underlying reward-driven task, independent of safety constraints. We demonstrate that this approach both ensures safety and effectively guides exploration during training in a range of experiments, including zero-shot transfer when the reward is learned in a modular way.
Driving safely requires multiple capabilities from human and intelligent agents, such as the generalizability to unseen environments, the safety awareness of the surrounding traffic, and the decision-making in complex multi-agent settings. Despite the great success of Reinforcement Learning (RL), most of the RL research works investigate each capability separately due to the lack of integrated environments. In this work, we develop a new driving simulation platform called MetaDrive to support the research of generalizable reinforcement learning algorithms for machine autonomy. MetaDrive is highly compositional, which can generate an infinite number of diverse driving scenarios from both the procedural generation and the real data importing. Based on MetaDrive, we construct a variety of RL tasks and baselines in both single-agent and multi-agent settings, including benchmarking generalizability across unseen scenes, safe exploration, and learning multi-agent traffic. The generalization experiments conducted on both procedurally generated scenarios and real-world scenarios show that increasing the diversity and the size of the training set leads to the improvement of the RL agent's generalizability. We further evaluate various safe reinforcement learning and multi-agent reinforcement learning algorithms in MetaDrive environments and provide the benchmarks. Source code, documentation, and demo video are available at \url{ https://metadriverse.github.io/metadrive}.
The last half-decade has seen a steep rise in the number of contributions on safe learning methods for real-world robotic deployments from both the control and reinforcement learning communities. This article provides a concise but holistic review of the recent advances made in using machine learning to achieve safe decision making under uncertainties, with a focus on unifying the language and frameworks used in control theory and reinforcement learning research. Our review includes: learning-based control approaches that safely improve performance by learning the uncertain dynamics, reinforcement learning approaches that encourage safety or robustness, and methods that can formally certify the safety of a learned control policy. As data- and learning-based robot control methods continue to gain traction, researchers must understand when and how to best leverage them in real-world scenarios where safety is imperative, such as when operating in close proximity to humans. We highlight some of the open challenges that will drive the field of robot learning in the coming years, and emphasize the need for realistic physics-based benchmarks to facilitate fair comparisons between control and reinforcement learning approaches.
Training-time safety violations have been a major concern when we deploy reinforcement learning algorithms in the real world. This paper explores the possibility of safe RL algorithms with zero training-time safety violations in the challenging setting where we are only given a safe but trivial-reward initial policy without any prior knowledge of the dynamics model and additional offline data. We propose an algorithm, Co-trained Barrier Certificate for Safe RL (CRABS), which iteratively learns barrier certificates, dynamics models, and policies. The barrier certificates, learned via adversarial training, ensure the policy's safety assuming calibrated learned dynamics model. We also add a regularization term to encourage larger certified regions to enable better exploration. Empirical simulations show that zero safety violations are already challenging for a suite of simple environments with only 2-4 dimensional state space, especially if high-reward policies have to visit regions near the safety boundary. Prior methods require hundreds of violations to achieve decent rewards on these tasks, whereas our proposed algorithms incur zero violations.
Context: Machine Learning (ML) has been at the heart of many innovations over the past years. However, including it in so-called 'safety-critical' systems such as automotive or aeronautic has proven to be very challenging, since the shift in paradigm that ML brings completely changes traditional certification approaches. Objective: This paper aims to elucidate challenges related to the certification of ML-based safety-critical systems, as well as the solutions that are proposed in the literature to tackle them, answering the question 'How to Certify Machine Learning Based Safety-critical Systems?'. Method: We conduct a Systematic Literature Review (SLR) of research papers published between 2015 to 2020, covering topics related to the certification of ML systems. In total, we identified 217 papers covering topics considered to be the main pillars of ML certification: Robustness, Uncertainty, Explainability, Verification, Safe Reinforcement Learning, and Direct Certification. We analyzed the main trends and problems of each sub-field and provided summaries of the papers extracted. Results: The SLR results highlighted the enthusiasm of the community for this subject, as well as the lack of diversity in terms of datasets and type of models. It also emphasized the need to further develop connections between academia and industries to deepen the domain study. Finally, it also illustrated the necessity to build connections between the above mention main pillars that are for now mainly studied separately. Conclusion: We highlighted current efforts deployed to enable the certification of ML based software systems, and discuss some future research directions.
Many sequential decision problems involve finding a policy that maximizes total reward while obeying safety constraints. Although much recent research has focused on the development of safe reinforcement learning (RL) algorithms that produce a safe policy after training, ensuring safety during training as well remains an open problem. A fundamental challenge is performing exploration while still satisfying constraints in an unknown Markov decision process (MDP). In this work, we address this problem for the chance-constrained setting. We propose a new algorithm, SAILR, that uses an intervention mechanism based on advantage functions to keep the agent safe throughout training and optimizes the agent's policy using off-the-shelf RL algorithms designed for unconstrained MDPs. Our method comes with strong guarantees on safety during both training and deployment (i.e., after training and without the intervention mechanism) and policy performance compared to the optimal safety-constrained policy. In our experiments, we show that SAILR violates constraints far less during training than standard safe RL and constrained MDP approaches and converges to a well-performing policy that can be deployed safely without intervention. Our code is available at https://github.com/nolanwagener/safe_rl.
The safety constraints commonly used by existing safe reinforcement learning (RL) methods are defined only on expectation of initial states, but allow each certain state to be unsafe, which is unsatisfying for real-world safety-critical tasks. In this paper, we introduce the feasible actor-critic (FAC) algorithm, which is the first model-free constrained RL method that considers statewise safety, e.g, safety for each initial state. We claim that some states are inherently unsafe no matter what policy we choose, while for other states there exist policies ensuring safety, where we say such states and policies are feasible. By constructing a statewise Lagrange function available on RL sampling and adopting an additional neural network to approximate the statewise Lagrange multiplier, we manage to obtain the optimal feasible policy which ensures safety for each feasible state and the safest possible policy for infeasible states. Furthermore, the trained multiplier net can indicate whether a given state is feasible or not through the statewise complementary slackness condition. We provide theoretical guarantees that FAC outperforms previous expectation-based constrained RL methods in terms of both constraint satisfaction and reward optimization. Experimental results on both robot locomotive tasks and safe exploration tasks verify the safety enhancement and feasibility interpretation of the proposed method.
Model information can be used to predict future trajectories, so it has huge potential to avoid dangerous region when implementing reinforcement learning (RL) on real-world tasks, like autonomous driving. However, existing studies mostly use model-free constrained RL, which causes inevitable constraint violations. This paper proposes a model-based feasibility enhancement technique of constrained RL, which enhances the feasibility of policy using generalized control barrier function (GCBF) defined on the distance to constraint boundary. By using the model information, the policy can be optimized safely without violating actual safety constraints, and the sample efficiency is increased. The major difficulty of infeasibility in solving the constrained policy gradient is handled by an adaptive coefficient mechanism. We evaluate the proposed method in both simulations and real vehicle experiments in a complex autonomous driving collision avoidance task. The proposed method achieves up to four times fewer constraint violations and converges 3.36 times faster than baseline constrained RL approaches.
Sepsis is a potentially life threatening inflammatory response to infection or severe tissue damage. It has a highly variable clinical course, requiring constant monitoring of the patient's state to guide the management of intravenous fluids and vasopressors, among other interventions. Despite decades of research, there's still debate among experts on optimal treatment. Here, we combine for the first time, distributional deep reinforcement learning with mechanistic physiological models to find personalized sepsis treatment strategies. Our method handles partial observability by leveraging known cardiovascular physiology, introducing a novel physiology-driven recurrent autoencoder, and quantifies the uncertainty of its own results. Moreover, we introduce a framework for uncertainty aware decision support with humans in the loop. We show that our method learns physiologically explainable, robust policies that are consistent with clinical knowledge. Further our method consistently identifies high risk states that lead to death, which could potentially benefit from more frequent vasopressor administration, providing valuable guidance for future research
This paper studies the constrained/safe reinforcement learning (RL) problem with sparse indicator signals for constraint violations. We propose a model-based approach to enable RL agents to effectively explore the environment with unknown system dynamics and environment constraints given a significantly small number of violation budgets. We employ the neural network ensemble model to estimate the prediction uncertainty and use model predictive control as the basic control framework. We propose the robust cross-entropy method to optimize the control sequence considering the model uncertainty and constraints. We evaluate our methods in the Safety Gym environment. The results show that our approach learns to complete the tasks with a much smaller number of constraint violations than state-of-the-art baselines. Additionally, we are able to achieve several orders of magnitude better sample efficiency when compared with constrained model-free RL approaches. The code is available at \url{https://github.com/liuzuxin/safe-mbrl}.
Safe reinforcement learning has been a promising approach for optimizing the policy of an agent that operates in safety-critical applications. In this paper, we propose an algorithm, SNO-MDP, that explores and optimizes Markov decision processes under unknown safety constraints. Specifically, we take a stepwise approach for optimizing safety and cumulative reward. In our method, the agent first learns safety constraints by expanding the safe region, and then optimizes the cumulative reward in the certified safe region. We provide theoretical guarantees on both the satisfaction of the safety constraint and the near-optimality of the cumulative reward under proper regularity assumptions. In our experiments, we demonstrate the effectiveness of SNO-MDP through two experiments: one uses a synthetic data in a new, openly-available environment named GP-SAFETY-GYM, and the other simulates Mars surface exploration by using real observation data.
In safety-critical applications, autonomous agents may need to learn in an environment where mistakes can be very costly. In such settings, the agent needs to behave safely not only after but also while learning. To achieve this, existing safe reinforcement learning methods make an agent rely on priors that let it avoid dangerous situations during exploration with high probability, but both the probabilistic guarantees and the smoothness assumptions inherent in the priors are not viable in many scenarios of interest such as autonomous driving. This paper presents an alternative approach inspired by human teaching, where an agent learns under the supervision of an automatic instructor that saves the agent from violating constraints during learning. In this model, we introduce the monitor that neither needs to know how to do well at the task the agent is learning nor needs to know how the environment works. Instead, it has a library of reset controllers that it activates when the agent starts behaving dangerously, preventing it from doing damage. Crucially, the choices of which reset controller to apply in which situation affect the speed of agent learning. Based on observing agents' progress, the teacher itself learns a policy for choosing the reset controllers, a curriculum, to optimize the agent's final policy reward. Our experiments use this framework in two environments to induce curricula for safe and efficient learning.
We study the safe reinforcement learning problem with nonlinear function approximation, where policy optimization is formulated as a constrained optimization problem with both the objective and the constraint being nonconvex functions. For such a problem, we construct a sequence of surrogate convex constrained optimization problems by replacing the nonconvex functions locally with convex quadratic functions obtained from policy gradient estimators. We prove that the solutions to these surrogate problems converge to a stationary point of the original nonconvex problem. Furthermore, to extend our theoretical results, we apply our algorithm to examples of optimal control and multi-agent reinforcement learning with safety constraints.
Reinforcement Learning (RL) has emerged as an efficient method of choice for solving complex sequential decision making problems in automatic control, computer science, economics, and biology. In this paper we present a model-free RL algorithm to synthesize control policies that maximize the probability of satisfying high-level control objectives given as Linear Temporal Logic (LTL) formulas. Uncertainty is considered in the workspace properties, the structure of the workspace, and the agent actions, giving rise to a Probabilistically-Labeled Markov Decision Process (PL-MDP) with unknown graph structure and stochastic behaviour, which is even more general case than a fully unknown MDP. We first translate the LTL specification into a Limit Deterministic Buchi Automaton (LDBA), which is then used in an on-the-fly product with the PL-MDP. Thereafter, we define a synchronous reward function based on the acceptance condition of the LDBA. Finally, we show that the RL algorithm delivers a policy that maximizes the satisfaction probability asymptotically. We provide experimental results that showcase the efficiency of the proposed method.
Navigating urban environments represents a complex task for automated vehicles. They must reach their goal safely and efficiently while considering a multitude of traffic participants. We propose a modular decision making algorithm to autonomously navigate intersections, addressing challenges of existing rule-based and reinforcement learning (RL) approaches. We first present a safe RL algorithm relying on a model-checker to ensure safety guarantees. To make the decision strategy robust to perception errors and occlusions, we introduce a belief update technique using a learning based approach. Finally, we use a scene decomposition approach to scale our algorithm to environments with multiple traffic participants. We empirically demonstrate that our algorithm outperforms rule-based methods and reinforcement learning techniques on a complex intersection scenario.
Reinforcement Learning (RL) is a widely employed machine learning architecture that has been applied to a variety of control problems. However, applications in safety-critical domains require a systematic and formal approach to specifying requirements as tasks or goals. We propose a model-free RL algorithm that enables the use of Linear Temporal Logic (LTL) to formulate a goal for unknown continuous-state/action Markov Decision Processes (MDPs). The given LTL property is translated into a Limit-Deterministic Generalised Buchi Automaton (LDGBA), which is then used to shape a synchronous reward function on-the-fly. Under certain assumptions, the algorithm is guaranteed to synthesise a control policy whose traces satisfy the LTL specification with maximal probability.
AI systems are increasingly applied to complex tasks that involve interaction with humans. During training, such systems are potentially dangerous, as they haven't yet learned to avoid actions that could cause serious harm. How can an AI system explore and learn without making a single mistake that harms humans or otherwise causes serious damage? For model-free reinforcement learning, having a human "in the loop" and ready to intervene is currently the only way to prevent all catastrophes. We formalize human intervention for RL and show how to reduce the human labor required by training a supervised learner to imitate the human's intervention decisions. We evaluate this scheme on Atari games, with a Deep RL agent being overseen by a human for four hours. When the class of catastrophes is simple, we are able to prevent all catastrophes without affecting the agent's learning (whereas an RL baseline fails due to catastrophic forgetting). However, this scheme is less successful when catastrophes are more complex: it reduces but does not eliminate catastrophes and the supervised learner fails on adversarial examples found by the agent. Extrapolating to more challenging environments, we show that our implementation would not scale (due to the infeasible amount of human labor required). We outline extensions of the scheme that are necessary if we are to train model-free agents without a single catastrophe.