control barrier filtering

This study addresses the challenge of providing certifiable runtime safety guarantees prior to tool invocation, focusing on three core issues: the representability of policy states, the observability of monitoring evidence, and the impact of interventions on future behavior. To this end, we propose the first formal theoretical framework for runtime safety-executable boundaries, distinguishing among static policy executability, statistical calibration under exogenous legal constraints, and closed-loop intervention effects. Building upon finitely controlled models, we develop a method for closed-loop safety certification that integrates register model identification, Neyman–Pearson hypothesis testing, conformal calibration, and occupancy planning. Empirical validation through static diagnosis, model enumeration, representation rewriting, and closed-loop re-execution experiments demonstrates the efficacy of our approach and exposes the fundamental limitations of static calibration under representation attacks.

This work addresses the lack of formal collision-avoidance guarantees for learning-based motion planners in complex road environments. We propose a real-time safety filter grounded in Control Barrier Functions (CBFs), enabling rigorous safety enforcement without compromising planning fidelity. Our method is the first to embed exact, non-conservative safety constraints for arbitrarily shaped road boundaries—eliminating the need for geometric approximations. Safety is enforced via online minimal-intervention quadratic programming (QP), which rectifies control commands while preserving the planner’s original intent. Extensive evaluation across challenging scenarios—including high-curvature, multi-branch, and narrow road segments—demonstrates 100% safety compliance and an average computation frequency of 40 Hz. The implementation, including source code and demonstration videos, is publicly available.

This work addresses the challenge of runtime safety certification under dynamic constraints in unstructured environments by proposing a Policy Library Control Barrier Function (PL-CBF) framework. The approach establishes a finite-horizon rollback mechanism that operates multiple policies in parallel, leverages language metric theory to formally characterize policy library coverage conditions, and employs quadratic programming to minimally modify a nominal policy by selecting the least intrusive safe mode. Experimental evaluations on double integrator systems, nonlinear vehicle models, and 3D quadrotor platforms demonstrate that PL-CBF substantially outperforms conventional single-policy filtering methods, achieving higher safety coverage while maintaining millisecond-level computational efficiency.

This paper addresses the challenge of simultaneously ensuring ω-regular safety (avoiding undesirable events) and liveness (guaranteeing eventual occurrence of desirable events) in learned probabilistic policies during runtime. To this end, we propose STARs, a dynamic runtime shielding framework. STARs is the first to support dynamic post-hoc shielding for the full class of ω-regular properties, leveraging policy templates, ω-automaton construction, game-theoretic model checking, and real-time monitoring with reconfiguration. It enables specification evolution and adaptive intervention under actuator failures. A key innovation is a tunable intervention mechanism that dynamically balances formal assurance strength against task performance. Evaluated on a mobile robot benchmark, STARs demonstrates low-overhead, highly controllable shielding for incrementally updated ω-regular specifications—significantly enhancing both practicality and trustworthiness of learned policies.

Reinforcement learning (RL) often neglects safety considerations, while online safety filters—such as control barrier functions (CBFs)—tend to induce overly conservative policies. Method: This paper introduces CBF-RL, the first framework that explicitly embeds CBF-based safety constraints into the RL training process, enabling the policy to autonomously internalize safe behavior during learning and eliminating runtime dependence on online safety filters. The method ensures closed-loop safety guarantees in discrete-time settings, enhancing both exploration safety and convergence speed. Contribution/Results: Evaluated on simulated navigation tasks and the Unitree G1 humanoid robot, CBF-RL achieves stable obstacle avoidance and stair climbing without online filtering. It significantly improves training efficiency and robustness under uncertainty, demonstrating superior performance over conventional RL and filtered baselines.

This work addresses the limitation of traditional safety mechanisms, which rely solely on state predicates and struggle to enforce smoothness constraints on higher-order derivatives such as velocity and acceleration. To overcome this, the paper proposes a higher-order safety shield synthesis method grounded in finite-state safety games. By employing finite differences to formally encode differential safety properties within a discrete state space, the problem is transformed into a safety game requiring memory of past states. The authors theoretically establish that enforcing a k-th order property necessitates preserving exactly k steps of history, and leverage this insight to design a layered, iterative algorithm for synthesizing maximally permissive strategies. This approach substantially reduces the search space, enhances synthesis efficiency, and effectively supports runtime enforcement of multi-order physical constraints.

This work addresses the gap between theoretical safety guarantees and practical feasibility of Control Barrier Functions (CBFs) in real-world systems subject to input constraints, where implicit assumptions often render CBFs ineffective. By systematically distinguishing between candidate and valid CBFs, the study uncovers the true source of safety in passive systems and extends safety verification to non-passive systems. Integrating system dynamics, explicit input constraint modeling, and class-K function analysis, the authors establish precise conditions under which CBFs yield valid safety assurances in low-dimensional systems and derive actionable design principles for safe controllers. An accompanying interactive web platform visually illustrates the core mechanisms and common pitfalls, offering practitioners an intuitive guide for reliable deployment.

This work addresses the challenge of balancing performance and safety constraints in model-free reinforcement learning when an accurate dynamics model is unavailable. The authors propose a robust Koopman-CBF framework that constructs affine control barrier functions (CBFs) in a lifted space using a finite-dimensional Koopman operator and employs a quadratic programming-based safety layer to correct policy actions in real time. To mitigate errors from Koopman approximation, they introduce a projection residual margin derived from historical trajectories and design an actor regularization mechanism to reduce reliance on the safety filter. Experiments demonstrate that the method achieves zero constraint violations in the CartPole task while matching the return of unconstrained SAC, and significantly reduces violations in high-dimensional Safety Gymnasium tasks. The study also reveals limitations of first-order velocity barriers and linear EDMD models.

This work addresses the challenge of ensuring safe and autonomous robot navigation in complex dynamic environments by proposing a novel “Control Barrier Corridor” framework. It unifies control barrier functions with safety corridors for the first time, reformulating safety constraints as locally feasible target regions. By integrating feedback control with convex optimization, the method generates reference trajectories that guarantee continuous safety in real time. The approach is validated on fully actuated systems, unicycle models, and linear output regulation systems, demonstrating its broad applicability. A key contribution lies in establishing a tunable trade-off between safety and responsiveness, enabling verifiably safe, persistent, and adaptive exploration even in unknown environments.

This work addresses the challenge of providing reliable safety guarantees in high-dimensional systems, where learning-based safety filters often fail due to prediction errors. To overcome this limitation, the paper proposes Adaptive Conformal Filtering (ACoFi), a novel framework that integrates Hamilton-Jacobi reachability analysis with adaptive conformal inference to dynamically adjust the policy switching threshold in response to prediction uncertainty. ACoFi offers an asymptotically valid, user-specified upper bound on the miscoverage rate of safety predictions, thereby delivering a soft yet theoretically grounded safety guarantee. Experimental evaluations on Dubins car and Safety Gymnasium benchmarks demonstrate that ACoFi significantly outperforms fixed-threshold baselines, achieving higher safety performance and fewer constraint violations—particularly in out-of-distribution scenarios.