Learning-Based Measurement-Robust Control Barrier Functions for Obstacle Avoidance under State Estimation Error
arXiv:2608.20467
2026
Dynamics
2 ideas extracted · analyzed Sep 1, 2026
What the math gives to ML
The paper provides a constructive mechanism for enforcing safety when a controller receives an uncertain state estimate: augment the control-barrier condition with a worst-case optimization over the admissible estimation-error set. Its most transferable asset is the two-stage replacement of an expensive robust inner minimization by a neural surrogate trained on robust labels and then corrected using differentiable closed-loop rollouts. In neural networks, this can become a measurement-robust output shield for learned policies, or a robustly trained safety critic whose violation boundary is explicitly tested against the certified uncertainty radius.
Ideas from this paper
✗ Failed on benchmark
2026
Distill the expensive inner minimization over state-estimation errors into a neural correction term that predicts the robust barrier drift, then fine-tune the correction using differentiable closed-loop rollouts. This retains the robustness mechanism while reducing the repeated optimization cost and allowing less conservative behavior than fixed analytic uncertainty bounds.
Useful8/10
Difficulty6/10
Novelty8/10
✗ Mechanism failed
2026
Attach a differentiable control-barrier safety filter to an RL or imitation policy when the policy observes an estimated state rather than the true state. The filter chooses the smallest correction to the network action that satisfies a barrier inequality for every state perturbation inside the known measurement-error set, preventing nominally safe actions from becoming unsafe after observation noise.
Useful8/10
Difficulty5/10
Novelty7/10