Golden Handcuffs make safer AI agents

Training agents directly on the task they will be deployed on often leads to the best policy performance, but training on physical robots can cause damage through falls, so minimizing training falls is critical. A standard mitigation hands control to a separate recovery policy whenever the agent leaves a designer-specified safe region. The resulting mixed-policy rollouts bias the on-policy update, and importance sampling cannot remove this bias when the recovery policy is deterministic. We propose SafeExplorer, a drop-in modification of proximal policy optimization (PPO) built on an unbiased policy-gradient estimator of the mixed-policy return. The estimator uses the score function only inside the safe region and never evaluates the recovery policy's density, so deterministic recovery policies pose no problem. On final reward, the estimator matches or outperforms truncated importance sampling, which is itself biased for that same gradient. SafeExplorer also speeds up learning near the safe-region boundary with a closed-form critic target at recovery-triggering states when dynamics and recovery are deterministic, and an imitation loss that copies recovery actions only from successful recoveries. On three simulated locomotion environments with ten seeds each, and given an already available recovery controller, SafeExplorer incurs $255\times$, $43\times$, and $21\times$ fewer training falls than standard PPO on HalfCheetah, Ant, and Unitree Go1. SafeExplorer also matches or exceeds the final reward of PPO. On Ant, where 8.7% of recovery interventions end in a fall, over $40\times$ the HalfCheetah rate, SafeExplorer reaches the success threshold on every seed and none of the five baselines reaches the threshold on any seed.

Most learning algorithms with formal regret guarantees essentially rely on trying all possible behaviors, which is problematic when some errors cannot be recovered from. Instead, we allow the learning agent to ask for help from a mentor and to transfer knowledge between similar states. We show that this combination enables the agent to learn both safely and effectively. Under standard online learning assumptions, we provide an algorithm whose regret and number of mentor queries are both sublinear in the time horizon for Markov decision processes with irreversible dynamics and infinite state spaces. Our proof involves a sequence of three reductions, making our result more general than a single algorithm. Conceptually, our result may be the first formal proof that it is possible for an agent to obtain high reward while becoming self-sufficient in an unknown, unbounded, and high-stakes environment without resets.

An agent must try new behaviors to explore and improve. In high-stakes environments, an agent that violates safety constraints may cause harm and must be taken offline, curtailing any future interaction. Imitating old behavior is safe, but excessive conservatism discourages exploration. How much behavior change is too much? We show how to use any safe reference policy as a probabilistic regulator for any optimized but untested policy. Conformal calibration on data from the safe policy determines how aggressively the new policy can act, while provably enforcing the user's declared risk tolerance. Unlike conservative optimization methods, we do not assume the user has identified the correct model class nor tuned any hyperparameters. Unlike previous conformal methods, our theory provides finite-sample guarantees even for non-monotonic bounded loss functions, and it introduces a new policy control setting. Our experiments on applications ranging from natural language question answering to biomolecular engineering show that safe exploration is not only possible from the first moment of deployment, but can also improve performance.