ORN-CBF: Learning Observation-conditioned Residual Neural Control Barrier Functions via Hypernetworks
This work addresses safety-critical control for autonomous systems like ground robots and quadcopters, offering incremental improvements in learning-based CBFs.
The paper tackled the challenge of designing control barrier functions for safety-critical autonomous systems in partially observable environments by proposing observation-conditioned neural CBFs based on Hamilton-Jacobi reachability analysis, resulting in improved success rates and generalization to out-of-domain environments compared to baselines.
Control barrier functions (CBFs) have been demonstrated as an effective method for safety-critical control of autonomous systems. Although CBFs are simple to deploy, their design remains challenging, motivating the development of learning-based approaches. Yet, issues such as suboptimal safe sets, applicability in partially observable environments, and lack of rigorous safety guarantees persist. In this work, we propose observation-conditioned neural CBFs based on Hamilton-Jacobi (HJ) reachability analysis, which approximately recover the maximal safe sets. We exploit certain mathematical properties of the HJ value function, ensuring that the predicted safe set never intersects with the observed failure set. Moreover, we leverage a hypernetwork-based architecture that is particularly suitable for the design of observation-conditioned safety filters. The proposed method is examined both in simulation and hardware experiments for a ground robot and a quadcopter. The results show improved success rates and generalization to out-of-domain environments compared to the baselines.