Abstract:Using control barrier functions (CBFs) as safety filters provides a computationally inexpensive yet effective method for constructing controllers in safety-critical applications. However, using CBFs requires the construction of a valid CBF, which is well known to be a challenging task, and accurate system dynamics, which are often unavailable. This paper presents a learning-based approach to learn a valid CBF and the system dynamics starting from a conservative handcrafted CBF (HCBF) and the nominal system dynamics. We devise new loss functions that better suit the CBF refinement pipeline and are able to produce well-behaved CBFs with the usage of distance functions. By adopting an episodic learning approach, our proposed method is able to learn the system dynamics while not requiring additional interactions with the environment. Additionally, we provide a theoretical analysis of the quality of the learned system dynamics. We show that our proposed learning approach can effectively learn a valid CBF and an estimation of the actual system dynamics. The effectiveness of our proposed method is empirically demonstrated through simulation studies on three systems, a double integrator, a unicycle, and a two-link arm.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to construct an effective control barrier function (CBF) and a system dynamics model through learning methods under uncertain dynamics conditions. Specifically, the paper focuses on two main challenges: 1. **Constructing an effective CBF**: Using CBF as a safety filter provides an effective and computationally inexpensive method for constructing controllers in critical applications. However, constructing an effective CBF requires an accurate system dynamics model, which is often difficult to obtain in practical applications. Moreover, finding an effective CBF is itself a challenging task, especially when the constraint conditions are nonlinear or non - convex. 2. **Handling uncertain dynamics**: In many practical applications, the system's dynamics model is often inaccurate because of unmodeled dynamics and parameter errors. Therefore, how to learn effective CBFs and dynamics models under these uncertainties is an important issue. To solve these problems, the paper proposes a learning - based method, starting from a conservative hand - designed CBF (HCBF) and the nominal system dynamics, and gradually learning and optimizing the CBF and the system dynamics model. The main contributions of the paper include: 1. **Learning a better CBF from HCBF**: By introducing a distance function as prior knowledge, a method is developed to learn a well - behaved CBF that can recover more parts of the safety set. 2. **Extending the CBF optimization problem to handle uncertain dynamics**: A new loss function is proposed, making the method still effective under uncertain dynamics conditions. 3. **Verifying the effectiveness of the method through extensive simulation studies**: Simulation studies are carried out on three different systems, including the double integrator, the unicycle, and the two - link manipulator, demonstrating the effectiveness of the proposed method. Through these contributions, the paper aims to provide a method that can effectively learn and optimize CBFs under uncertain dynamics conditions, thereby improving the safety of safety - critical systems.

Learning a Better Control Barrier Function Under Uncertain Dynamics

Whole-body Dynamic Collision Avoidance with Time-varying Control Barrier Functions

Learning Performance-Oriented Control Barrier Functions Under Complex Safety Constraints and Limited Actuation

Learning-based Parameterized Barrier Function for Safety-Critical Control of Unknown Systems

Learning for Safety-Critical Control with Control Barrier Functions

Learning Local Control Barrier Functions for Hybrid Systems

State-action control barrier functions: Imposing safety on learning-based control with low online computational costs

Learning Control Barrier Functions with High Relative Degree for Safety-Critical Control

Control Barriers in Bayesian Learning of System Dynamics

Robust Safe Learning and Control in An Unknown Environment: An Uncertainty-Separated Control Barrier Function Approach

Safe and Stable Adaptive Control for a Class of Dynamic Systems

Learning Differentiable Safety-Critical Control using Control Barrier Functions for Generalization to Novel Environments

On the Optimality, Stability, and Feasibility of Control Barrier Functions: An Adaptive Learning-Based Approach

Transfer of Safety Controllers Through Learning Deep Inverse Dynamics Model

Learning-Based Control Barrier Function with Provably Safe Guarantees: Reducing Conservatism with Heading-Aware Safety Margin

Fast, Smooth, and Safe: Implicit Control Barrier Functions through Reach-Avoid Differential Dynamic Programming

Safe Online Dynamics Learning with Initially Unknown Models and Infeasible Safety Certificates

Safety-Aware Learning-Based Control of Systems with Uncertainty Dependent Constraints (extended version)

Learning for Layered Safety-Critical Control with Predictive Control Barrier Functions