Abstract:In this work, we investigate the data-driven safe control synthesis problem for unknown dynamic systems. We first formulate the safety synthesis problem as a robust convex program (RCP) based on notion of control barrier function. To resolve the issue of unknown system dynamic, we follow the existing approach by converting the RCP to a scenario convex program (SCP) by randomly collecting finite samples of system trajectory. However, to improve the sample efficiency to achieve a desired confidence bound, we provide a new posteriori method with validation tests. Specifically, after collecting a set of data for the SCP, we further collect another set of independent \emph{validate data} as posterior information to test the obtained solution. We derive a new overall confidence bound for the safety of the controller that connects the original sample data, the support constraints, and the validation data. The efficiency of the proposed approach is illustrated by a case study of room temperature control. We show that, compared with existing methods, the proposed approach can significantly reduce the required number of sample data to achieve a desired confidence bound.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop a data - driven method for synthesizing safety controllers for unknown dynamic systems. Specifically, researchers are concerned with how to ensure the safety of the controller through a limited number of data samples without an accurate system model, and minimize the number of required samples to achieve the expected confidence level. ### Main Problem Description The research problem proposed in the paper is for an unknown discrete - time deterministic system (dt - DS), using a data - driven method to design a controller \(C\) such that the closed - loop system \(S_C\) satisfies the given safety property \(\phi=(X_0, X_u, T)\), where: - \(X_0\subseteq X\) represents the initial region, - \(X_u\subseteq X\) represents the unsafe region, - \(T\) represents the time range of the property. The goal is to use a data - driven method to find a controller \(C: X\rightarrow U\) such that \(S_C\models\phi\), that is, for all initial states \(x(0)\in X_0\), the trajectory \(\zeta_T(x(0))\) of the system will not enter the unsafe region \(X_u\) within the time range \(T\), and ensure that this conclusion holds with a confidence level of \((1 - \beta)\). ### Limitations of Existing Methods Existing methods mainly rely on control barrier functions (CBF) and scenario convex programming (SCP), but these methods face the following challenges when dealing with unknown dynamic systems: 1. **Low sample efficiency**: A large amount of sample data is required to achieve a certain confidence level. 2. **Lack of verification mechanism**: Existing methods are usually only optimized based on the original sample data, without an effective verification mechanism to improve the confidence level. ### Innovations of the Paper To solve the above problems, this paper introduces a new posteriori method with validation tests, whose main features include: 1. **Two - stage data collection**: First, collect a set of data for constructing scenario convex programming (SCP), and then collect another set of independent verification data as posterior information for testing the obtained solution. 2. **Joint use of support constraints and verification frequency**: Not only consider the number of support constraints, but also combine the violation frequency in the verification data, thereby improving sample efficiency and confidence level. 3. **New overall performance bound**: Derive a new overall performance bound that combines the information of the original sample data, support constraints, and verification data in a unified manner, providing a more stringent performance guarantee. ### Experimental Verification Through a case study of room temperature control, the paper demonstrates the effectiveness of the proposed method. Compared with existing methods, the new method significantly reduces the number of required samples while achieving the same confidence level. ### Summary The main contribution of this paper is to propose a more efficient data - driven safety controller synthesis method. By introducing a posteriori verification mechanism, it effectively reduces the sample requirements and improves the practicality and reliability of the algorithm.

Data-Driven Safe Controller Synthesis for Deterministic Systems: A Posteriori Method With Validation Tests

Synthesis and verification of robust-adaptive safe controllers

Towards Robust Data-Driven Control Synthesis for Nonlinear Systems with Actuation Uncertainty

Robust Data-Driven Safe Control using Density Functions

Safety-Critical Control Synthesis for Unknown Sampled-Data Systems via Control Barrier Functions

Data-driven abstraction-based control synthesis

Safe and Stable Control Synthesis for Uncertain System Models via Distributionally Robust Optimization

From a Single Trajectory to Safety Controller Synthesis of Discrete-Time Nonlinear Polynomial Systems

Safe Control Synthesis Using Environmentally Robust Control Barrier Functions

A Multi-Fidelity Bayesian Approach to Safe Controller Design

Learning Robust Safety Controllers for Uncertain Input-Affine Polynomial Systems

Data-driven verification and synthesis of stochastic systems via barrier certificates

Adaptive safety-critical control for a class of nonlinear systems with parametric uncertainties: A control barrier function approach

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

Semi-Supervised Safe Visuomotor Policy Synthesis using Barrier Certificates

Formal Synthesis of Controllers for Safety-Critical Autonomous Systems: Developments and Challenges

Data-Driven Strategy Synthesis for Stochastic Systems with Unknown Nonlinear Disturbances

Robust Control for Dynamical Systems with Non-Gaussian Noise via Formal Abstractions

Bayesian Formal Synthesis of Unknown Systems via Robust Simulation Relations

Provable Reach-avoid Controllers Synthesis Based on Inner-approximating Controlled Reach-avoid Sets