Safe reinforcement learning (RL) agents accomplish given tasks while adhering to specific constraints. Employing constraints expressed via easily-understandable human language offers considerable potential for real-world applications due to its accessibility and non-reliance on domain expertise. Previous safe RL methods with natural language constraints typically adopt a recurrent neural network, which leads to limited capabilities when dealing with various forms of human language input. Furthermore, these methods often require a ground-truth cost function, necessitating domain expertise for the conversion of language constraints into a well-defined cost function that determines constraint violation. To address these issues, we proposes to use pre-trained language models (LM) to facilitate RL agents' comprehension of natural language constraints and allow them to infer costs for safe policy learning. Through the use of pre-trained LMs and the elimination of the need for a ground-truth cost, our method enhances safe policy learning under a diverse set of human-derived free-form natural language constraints. Experiments on grid-world navigation and robot control show that the proposed method can achieve strong performance while adhering to given constraints. The usage of pre-trained LMs allows our method to comprehend complicated constraints and learn safe policies without the need for ground-truth cost at any stage of training or evaluation. Extensive ablation studies are conducted to demonstrate the efficacy of each part of our method.
When deploying machine learning algorithms in the real world, guaranteeing safety is an essential asset. Existing safe learning approaches typically consider continuous variables, i.e., regression tasks. However, in practice, robotic systems are also subject to discrete, external environmental changes, e.g., having to carry objects of certain weights or operating on frozen, wet, or dry surfaces. Such influences can be modeled as discrete context variables. In the existing literature, such contexts are, if considered, mostly assumed to be known. In this work, we drop this assumption and show how we can perform safe learning when we cannot directly measure the context variables. To achieve this, we derive frequentist guarantees for multi-class classification, allowing us to estimate the current context from measurements. Further, we propose an approach for identifying contexts through experiments. We discuss under which conditions we can retain theoretical guarantees and demonstrate the applicability of our algorithm on a Furuta pendulum with camera measurements of different weights that serve as contexts.
Safety is an indispensable requirement for applying reinforcement learning (RL) to real problems. Although there has been a surge of safe RL algorithms proposed in recent years, most existing work typically 1) relies on receiving numeric safety feedback; 2) does not guarantee safety during the learning process; 3) limits the problem to a priori known, deterministic transition dynamics; and/or 4) assume the existence of a known safe policy for any states. Addressing the issues mentioned above, we thus propose Long-term Binaryfeedback Safe RL (LoBiSaRL), a safe RL algorithm for constrained Markov decision processes (CMDPs) with binary safety feedback and an unknown, stochastic state transition function. LoBiSaRL optimizes a policy to maximize rewards while guaranteeing a long-term safety that an agent executes only safe state-action pairs throughout each episode with high probability. Specifically, LoBiSaRL models the binary safety function via a generalized linear model (GLM) and conservatively takes only a safe action at every time step while inferring its effect on future safety under proper assumptions. Our theoretical results show that LoBiSaRL guarantees the long-term safety constraint, with high probability. Finally, our empirical results demonstrate that our algorithm is safer than existing methods without significantly compromising performance in terms of reward.
Online safe reinforcement learning (RL) involves training a policy that maximizes task efficiency while satisfying constraints via interacting with the environments. In this paper, our focus lies in addressing the complex challenges associated with solving multi-constraint (MC) safe RL problems. We approach the safe RL problem from the perspective of Multi-Objective Optimization (MOO) and propose a unified framework designed for MC safe RL algorithms. This framework highlights the manipulation of gradients derived from constraints. Leveraging insights from this framework and recognizing the significance of \textit{redundant} and \textit{conflicting} constraint conditions, we introduce the Gradient Shaping (GradS) method for general Lagrangian-based safe RL algorithms to improve the training efficiency in terms of both reward and constraint satisfaction. Our extensive experimentation demonstrates the effectiveness of our proposed method in encouraging exploration and learning a policy that improves both safety and reward performance across various challenging MC safe RL tasks as well as good scalability to the number of constraints.
This paper studies safe Reinforcement Learning (safe RL) with linear function approximation and under hard instantaneous constraints where unsafe actions must be avoided at each step. Existing studies have considered safe RL with hard instantaneous constraints, but their approaches rely on several key assumptions: $(i)$ the RL agent knows a safe action set for {\it every} state or knows a {\it safe graph} in which all the state-action-state triples are safe, and $(ii)$ the constraint/cost functions are {\it linear}. In this paper, we consider safe RL with instantaneous hard constraints without assumption $(i)$ and generalize $(ii)$ to Reproducing Kernel Hilbert Space (RKHS). Our proposed algorithm, LSVI-AE, achieves $\tilde{\cO}(\sqrt{d^3H^4K})$ regret and $\tilde{\cO}(H \sqrt{dK})$ hard constraint violation when the cost function is linear and $\cO(H\gamma_K \sqrt{K})$ hard constraint violation when the cost function belongs to RKHS. Here $K$ is the learning horizon, $H$ is the length of each episode, and $\gamma_K$ is the information gain w.r.t the kernel used to approximate cost functions. Our results achieve the optimal dependency on the learning horizon $K$, matching the lower bound we provide in this paper and demonstrating the efficiency of LSVI-AE. Notably, the design of our approach encourages aggressive policy exploration, providing a unique perspective on safe RL with general cost functions and no prior knowledge of safe actions, which may be of independent interest.
A popular framework for enforcing safe actions in Reinforcement Learning (RL) is Constrained RL, where trajectory based constraints on expected cost (or other cost measures) are employed to enforce safety and more importantly these constraints are enforced while maximizing expected reward. Most recent approaches for solving Constrained RL convert the trajectory based cost constraint into a surrogate problem that can be solved using minor modifications to RL methods. A key drawback with such approaches is an over or underestimation of the cost constraint at each state. Therefore, we provide an approach that does not modify the trajectory based cost constraint and instead imitates ``good'' trajectories and avoids ``bad'' trajectories generated from incrementally improving policies. We employ an oracle that utilizes a reward threshold (which is varied with learning) and the overall cost constraint to label trajectories as ``good'' or ``bad''. A key advantage of our approach is that we are able to work from any starting policy or set of trajectories and improve on it. In an exhaustive set of experiments, we demonstrate that our approach is able to outperform top benchmark approaches for solving Constrained RL problems, with respect to expected cost, CVaR cost, or even unknown cost constraints.
This paper introduces the reinforcement learning backup shield (RLBUS), an algorithm that guarantees safe exploration in reinforcement learning (RL) by incorporating backup control barrier functions (BCBFs). RLBUS constructs an implicit control forward invariant subset of the safe set using multiple backup policies, ensuring safety in the presence of input constraints. While traditional BCBFs often result in conservative control forward-invariant sets due to the design of backup controllers, RLBUS addresses this limitation by leveraging model-free RL to train an additional backup policy, which enlarges the identified control forward invariant subset of the safe set. This approach enables the exploration of larger regions in the state space with zero safety violations during training. The effectiveness of RLBUS is demonstrated on an inverted pendulum example, where the expanded invariant set allows for safe exploration over a broader state space, enhancing performance without compromising safety.
Safe reinforcement learning deals with mitigating or avoiding unsafe situations by reinforcement learning (RL) agents. Safe RL approaches are based on specific risk representations for particular problems or domains. In order to analyze agent behaviors, compare safe RL approaches, and effectively transfer techniques between application domains, it is necessary to understand the types of risk specific to safe RL problems. We performed a systematic literature mapping with the objective to characterize risk in safe RL. Based on the obtained results, we present definitions, characteristics, and types of risk that hold on multiple application domains. Our literature mapping covers literature from the last 5 years (2017-2022), from a variety of knowledge areas (AI, finance, engineering, medicine) where RL approaches emphasize risk representation and management. Our mapping covers 72 papers filtered systematically from over thousands of papers on the topic. Our proposed notion of risk covers a variety of representations, disciplinary differences, common training exercises, and types of techniques. We encourage researchers to include explicit and detailed accounts of risk in future safe RL research reports, using this mapping as a starting point. With this information, researchers and practitioners could draw stronger conclusions on the effectiveness of techniques on different problems.
As safety is of paramount importance in robotics, reinforcement learning that reflects safety, called safe RL, has been studied extensively. In safe RL, we aim to find a policy which maximizes the desired return while satisfying the defined safety constraints. There are various types of constraints, among which constraints on conditional value at risk (CVaR) effectively lower the probability of failures caused by high costs since CVaR is a conditional expectation obtained above a certain percentile. In this paper, we propose a trust region-based safe RL method with CVaR constraints, called TRC. We first derive the upper bound on CVaR and then approximate the upper bound in a differentiable form in a trust region. Using this approximation, a subproblem to get policy gradients is formulated, and policies are trained by iteratively solving the subproblem. TRC is evaluated through safe navigation tasks in simulations with various robots and a sim-to-real environment with a Jackal robot from Clearpath. Compared to other safe RL methods, the performance is improved by 1.93 times while the constraints are satisfied in all experiments.
This paper delves into the problem of safe reinforcement learning (RL) in a partially observable environment with the aim of achieving safe-reachability objectives. In traditional partially observable Markov decision processes (POMDP), ensuring safety typically involves estimating the belief in latent states. However, accurately estimating an optimal Bayesian filter in POMDP to infer latent states from observations in a continuous state space poses a significant challenge, largely due to the intractable likelihood. To tackle this issue, we propose a stochastic model-based approach that guarantees RL safety almost surely in the face of unknown system dynamics and partial observation environments. We leveraged the Predictive State Representation (PSR) and Reproducing Kernel Hilbert Space (RKHS) to represent future multi-step observations analytically, and the results in this context are provable. Furthermore, we derived essential operators from the kernel Bayes' rule, enabling the recursive estimation of future observations using various operators. Under the assumption of \textit{undercompleness}, a polynomial sample complexity is established for the RL algorithm for the infinite size of observation and action spaces, ensuring an $\epsilon-$suboptimal safe policy guarantee.
This paper aims to solve a safe reinforcement learning (RL) problem with risk measure-based constraints. As risk measures, such as conditional value at risk (CVaR), focus on the tail distribution of cost signals, constraining risk measures can effectively prevent a failure in the worst case. An on-policy safe RL method, called TRC, deals with a CVaR-constrained RL problem using a trust region method and can generate policies with almost zero constraint violations with high returns. However, to achieve outstanding performance in complex environments and satisfy safety constraints quickly, RL methods are required to be sample efficient. To this end, we propose an off-policy safe RL method with CVaR constraints, called off-policy TRC. If off-policy data from replay buffers is directly used to train TRC, the estimation error caused by the distributional shift results in performance degradation. To resolve this issue, we propose novel surrogate functions, in which the effect of the distributional shift can be reduced, and introduce an adaptive trust-region constraint to ensure a policy not to deviate far from replay buffers. The proposed method has been evaluated in simulation and real-world environments and satisfied safety constraints within a few steps while achieving high returns even in complex robotic tasks.
This paper proposes a multiagent based bi-level operation framework for the low-carbon demand management in distribution networks considering the carbon emission allowance on the demand side. In the upper level, the aggregate load agents optimize the control signals for various types of loads to maximize the profits; in the lower level, the distribution network operator makes optimal dispatching decisions to minimize the operational costs and calculates the distribution locational marginal price and carbon intensity. The distributed flexible load agent has only incomplete information of the distribution network and cooperates with other agents using networked communication. Finally, the problem is formulated into a networked multi-agent constrained Markov decision process, which is solved using a safe reinforcement learning algorithm called consensus multi-agent constrained policy optimization considering the carbon emission allowance for each agent. Case studies with the IEEE 33-bus and 123-bus distribution network systems demonstrate the effectiveness of the proposed approach, in terms of satisfying the carbon emission constraint on demand side, ensuring the safe operation of the distribution network and preserving privacy of both sides.
Highly automated assembly lines enable significant productivity gains in the manufacturing industry, particularly in mass production condition. Nonetheless, challenges persist in job scheduling for make-to-job and mass customization, necessitating further investigation to improve efficiency, reduce tardiness, promote safety and reliability. In this contribution, an advantage actor-critic based reinforcement learning method is proposed to address scheduling problems of distributed flexible assembly lines in a real-time manner. To enhance the performance, a more condensed environment representation approach is proposed, which is designed to work with the masks made by priority dispatching rules to generate fixed and advantageous action space. Moreover, a Monte-Carlo tree search based soft shielding component is developed to help address long-sequence dependent unsafe behaviors and monitor the risk of overdue scheduling. Finally, the proposed algorithm and its soft shielding component are validated in performance evaluation.
In the context of safe exploration, Reinforcement Learning (RL) has long grappled with the challenges of balancing the tradeoff between maximizing rewards and minimizing safety violations, particularly in complex environments with contact-rich or non-smooth dynamics, and when dealing with high-dimensional pixel observations. Furthermore, incorporating state-wise safety constraints in the exploration and learning process, where the agent must avoid unsafe regions without prior knowledge, adds another layer of complexity. In this paper, we propose a novel pixel-observation safe RL algorithm that efficiently encodes state-wise safety constraints with unknown hazard regions through a newly introduced latent barrier-like function learning mechanism. As a joint learning framework, our approach begins by constructing a latent dynamics model with low-dimensional latent spaces derived from pixel observations. We then build and learn a latent barrier-like function on top of the latent dynamics and conduct policy optimization simultaneously, thereby improving both safety and the total expected return. Experimental evaluations on the safety-gym benchmark suite demonstrate that our proposed method significantly reduces safety violations throughout the training process, and demonstrates faster safety convergence compared to existing methods while achieving competitive results in reward return.
Incorporating safety is an essential prerequisite for broadening the practical applications of reinforcement learning in real-world scenarios. To tackle this challenge, Constrained Markov Decision Processes (CMDPs) are leveraged, which introduce a distinct cost function representing safety violations. In CMDPs' settings, Lagrangian relaxation technique has been employed in previous algorithms to convert constrained optimization problems into unconstrained dual problems. However, these algorithms may inaccurately predict unsafe behavior, resulting in instability while learning the Lagrange multiplier. This study introduces a novel safe reinforcement learning algorithm, Safety Critic Policy Optimization (SCPO). In this study, we define the safety critic, a mechanism that nullifies rewards obtained through violating safety constraints. Furthermore, our theoretical analysis indicates that the proposed algorithm can automatically balance the trade-off between adhering to safety constraints and maximizing rewards. The effectiveness of the SCPO algorithm is empirically validated by benchmarking it against strong baselines.
The difficulty of identifying the physical model of complex systems has led to exploring methods that do not rely on such complex modeling of the systems. Deep reinforcement learning has been the pioneer for solving this problem without the need for relying on the physical model of complex systems by just interacting with it. However, it uses a black-box learning approach that makes it difficult to be applied within real-world and safety-critical systems without providing explanations of the actions derived by the model. Furthermore, an open research question in deep reinforcement learning is how to focus the policy learning of critical decisions within a sparse domain. This paper proposes a novel approach for the use of deep reinforcement learning in safety-critical systems. It combines the advantages of probabilistic modeling and reinforcement learning with the added benefits of interpretability and works in collaboration and synchronization with conventional decision-making strategies. The BC-SRLA is activated in specific situations which are identified autonomously through the fused information of probabilistic model and reinforcement learning, such as abnormal conditions or when the system is near-to-failure. Further, it is initialized with a baseline policy using policy cloning to allow minimum interactions with the environment to address the challenges associated with using RL in safety-critical industries. The effectiveness of the BC-SRLA is demonstrated through a case study in maintenance applied to turbofan engines, where it shows superior performance to the prior art and other baselines.
Traditional controllers have limitations as they rely on prior knowledge about the physics of the problem, require modeling of dynamics, and struggle to adapt to abnormal situations. Deep reinforcement learning has the potential to address these problems by learning optimal control policies through exploration in an environment. For safety-critical environments, it is impractical to explore randomly, and replacing conventional controllers with black-box models is also undesirable. Also, it is expensive in continuous state and action spaces, unless the search space is constrained. To address these challenges we propose a specialized deep residual policy safe reinforcement learning with a cycle of learning approach adapted for complex and continuous state-action spaces. Residual policy learning allows learning a hybrid control architecture where the reinforcement learning agent acts in synchronous collaboration with the conventional controller. The cycle of learning initiates the policy through the expert trajectory and guides the exploration around it. Further, the specialization through the input-output hidden Markov model helps to optimize policy that lies within the region of interest (such as abnormality), where the reinforcement learning agent is required and is activated. The proposed solution is validated on the Tennessee Eastman process control.
With the development of large language models (LLMs), striking a balance between the performance and safety of AI systems has never been more critical. However, the inherent tension between the objectives of helpfulness and harmlessness presents a significant challenge during LLM training. To address this issue, we propose Safe Reinforcement Learning from Human Feedback (Safe RLHF), a novel algorithm for human value alignment. Safe RLHF explicitly decouples human preferences regarding helpfulness and harmlessness, effectively avoiding the crowdworkers' confusion about the tension and allowing us to train separate reward and cost models. We formalize the safety concern of LLMs as an optimization task of maximizing the reward function while satisfying specified cost constraints. Leveraging the Lagrangian method to solve this constrained problem, Safe RLHF dynamically adjusts the balance between the two objectives during fine-tuning. Through a three-round fine-tuning using Safe RLHF, we demonstrate a superior ability to mitigate harmful responses while enhancing model performance compared to existing value-aligned algorithms. Experimentally, we fine-tuned the Alpaca-7B using Safe RLHF and aligned it with collected human preferences, significantly improving its helpfulness and harmlessness according to human evaluations.
Artificial intelligence (AI) systems possess significant potential to drive societal progress. However, their deployment often faces obstacles due to substantial safety concerns. Safe reinforcement learning (SafeRL) emerges as a solution to optimize policies while simultaneously adhering to multiple constraints, thereby addressing the challenge of integrating reinforcement learning in safety-critical scenarios. In this paper, we present an environment suite called Safety-Gymnasium, which encompasses safety-critical tasks in both single and multi-agent scenarios, accepting vector and vision-only input. Additionally, we offer a library of algorithms named Safe Policy Optimization (SafePO), comprising 16 state-of-the-art SafeRL algorithms. This comprehensive library can serve as a validation tool for the research community. By introducing this benchmark, we aim to facilitate the evaluation and comparison of safety performance, thus fostering the development of reinforcement learning for safer, more reliable, and responsible real-world applications. The website of this project can be accessed at https://sites.google.com/view/safety-gymnasium.
A critical goal of autonomy and artificial intelligence is enabling autonomous robots to rapidly adapt in dynamic and uncertain environments. Classic adaptive control and safe control provide stability and safety guarantees but are limited to specific system classes. In contrast, policy adaptation based on reinforcement learning (RL) offers versatility and generalizability but presents safety and robustness challenges. We propose SafeDPA, a novel RL and control framework that simultaneously tackles the problems of policy adaptation and safe reinforcement learning. SafeDPA jointly learns adaptive policy and dynamics models in simulation, predicts environment configurations, and fine-tunes dynamics models with few-shot real-world data. A safety filter based on the Control Barrier Function (CBF) on top of the RL policy is introduced to ensure safety during real-world deployment. We provide theoretical safety guarantees of SafeDPA and show the robustness of SafeDPA against learning errors and extra perturbations. Comprehensive experiments on (1) classic control problems (Inverted Pendulum), (2) simulation benchmarks (Safety Gym), and (3) a real-world agile robotics platform (RC Car) demonstrate great superiority of SafeDPA in both safety and task performance, over state-of-the-art baselines. Particularly, SafeDPA demonstrates notable generalizability, achieving a 300% increase in safety rate compared to the baselines, under unseen disturbances in real-world experiments.