Safety is critical to broadening the application of reinforcement learning (RL). Often, we train RL agents in a controlled environment, such as a laboratory, before deploying them in the real world. However, the real-world target task might be unknown prior to deployment. Reward-free RL trains an agent without the reward to adapt quickly once the reward is revealed. We consider the constrained reward-free setting, where an agent (the guide) learns to explore safely without the reward signal. This agent is trained in a controlled environment, which allows unsafe interactions and still provides the safety signal. After the target task is revealed, safety violations are not allowed anymore. Thus, the guide is leveraged to compose a safe behaviour policy. Drawing from transfer learning, we also regularize a target policy (the student) towards the guide while the student is unreliable and gradually eliminate the influence of the guide as training progresses. The empirical analysis shows that this method can achieve safe transfer learning and helps the student solve the target task faster.@en

Safety is critical to broadening the a lication of reinforcement learning (RL). Often, RL agents are trained in a controlled environment, such as a laboratory, before being de loyed in the real world. However, the target reward might be unknown rior to de loyment. Reward-free RL addresses this roblem by training an agent without the reward to ada t quickly once the reward is revealed. We consider the constrained reward-free setting, where an agent (the guide) learns to ex lore safely without the reward signal. This agent is trained in a controlled environment, which allows unsafe interactions and still rovides the safety signal. After the target task is revealed, safety violations are not allowed anymore. Thus, the guide is leveraged to com ose a safe sam ling olicy. Drawing from transfer learning, we also regularize a target olicy (the student) towards the guide while the student is unreliable and gradually eliminate the influence from the guide as training rogresses. The em irical analysis shows that this method can achieve safe transfer learning and hel s the student solve the target task faster. @en

We study planning problems where a controllable agent operates under partial observability and interacts with an uncontrollable opponent, also referred to as the adversary. The agent has two distinct objectives: To maximize an expected value and to adhere to a safety specification. Multi-objective partially observable stochastic games (POSGs) formally model such problems. Yet, even for a single objective, the task of computing suitable policies for POSGs is theoretically hard and computationally intractable in practice. Using a factored state-space representation, we define a decoupling scheme for the POSG state space that—under certain assumptions on the observability and the reward structure—separates the state components relevant for the reward from those relevant for safety. This decoupling affects the possibility to compute provably safe and reward-optimal policies in a tractable two-stage approach. In particular, on the fully observable components related to safety, we exactly compute the set of policies that captures all possible safe choices against the opponent. We restrict the agent’s behavior to these safe policies and project the POSG to a partially observable Markov decision process (POMDP). Any reward-maximal policy for the POMDP is then guaranteed to be safe and reward-maximal for the POSG. We showcase our approach’s feasibility using high-fidelity simulations of two case studies that concern UAV path planning and autonomous driving. Moreover, to demonstrate the practical applicability, we design a physical experiment involving a robot decision making problem under energy constraints that is motivated by a paired helicopter with NASA’s Perseverance Mars rover.@en

Deploying reinforcement learning (RL) involves major concerns around safety. Engineering a reward signal that allows the agent to maximize its performance while remaining safe is not trivial. Safe RL studies how to mitigate such problems. For instance, we can decouple safety from reward using constrained Markov decision processes (CMDPs), where an independent signal models the safety aspects. In this setting, an RL agent can autonomously find tradeoffs between performance and safety. Unfortunately, most RL agents designed for CMDPs only guarantee safety after the learning phase, which might prevent their direct deployment. In this work, we investigate settings where a concise abstract model of the safety aspects is given, a reasonable assumption since a thorough understanding of safety-related matters is a prerequisite for deploying RL in typical applications. Factored CMDPs provide such compact models when a small subset of features describe the dynamics relevant for the safety constraints. We propose an RL algorithm that uses this abstract model to learn policies for CMDPs safely, that is without violating the constraints. During the training process, this algorithm can seamlessly switch from a conservative policy to a greedy policy without violating the safety constraints. We prove that this algorithm is safe under the given assumptions. Empirically, we show that even if safety and reward signals are contradictory, this algorithm always operates safely and, when they are aligned, this approach also improves the agent's performance.@en