This paper introduces Learning-Based Adaptive Control Barrier Functions (LB-A-CBFs) to address safety-critical control for systems with unknown nonlinear dynamics. The approach integrates a learning module based on neural operators, an adaptive parameter estimation mechanism, and a robust CBF formulation. Theoretical analysis and simulations demonstrate the effectiveness of the approach.
Key findings
Introduces LB-A-CBFs for safety-critical control in systems with unknown nonlinear dynamics.
Combines neural operator-based dynamics learning with adaptive parameter estimation.
Provides theoretical guarantees for input-to-state safety and convergence of the learning module.
Demonstrates effectiveness through simulations on complex robotic systems.
Limitations & open questions
The paper does not discuss the computational complexity of the proposed method.
The effectiveness of the approach is only demonstrated through simulations, not real-world experiments.