Synthesizing Min-Max Control Barrier Functions For Switched Affine Systems footnoteinfo

Sara Kamali, Guillaume O. Berger, Sriram Sankaranarayanan · 2025

Read Synthesizing Min-Max Control Barrier Functions For Switched Affine Systems footnoteinfo by Sara Kamali, Guillaume O. Berger, Sriram Sankaranarayanan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-10296.txt.

Introduction

Sara Kamali, Guillaume Berger, and Sriram Sankaranarayanan build safety certificates for systems that switch between simple modes, in a 2025 paper. Such switched systems pick their current mode by rule from their state, and a control barrier function is a test that keeps the state inside a safe set. The team studies certificates built as the smallest or largest of a set of flat functions, which lets them bend sharply.

Using tools from nonsmooth analysis, they state when such a certificate truly guarantees safety, and how to read a switching law off it. They then automate the search for certificates with a tree-search method borrowed from optimization. Tests on several switched systems show the approach finds working certificates.

Catalog record

CREATORSara Kamali, Guillaume O. Berger, Sriram Sankaranarayanan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE66 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-10296

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Synthesizing Min-Max Control Barrier Functions For Switched Affine Systems footnoteinfo Sara Kamali, Guillaume O. Berger, Sriram Sankaranarayanan Abstract We study the problem of synthesizing non-smooth control barrier functions (CBFs) for continuous-time switched affine systems. Switched affine systems are defined by a set of affine dynamical modes, wherein the control consists of a state-based switching signal that determines the current operating mode. The control barrier functions seek to maintain the system state inside a control invariant set that excludes a given set of unsafe states. We consider CBFs that take the form of pointwise minima and maxima over a finite set of affine functions. Our approach uses ideas from nonsmooth analysis to formulate conditions for min- and max- affine control barrier functions. We show how a feedback switching law can be extracted from a given CBF. Next, we show how to automate the process of synthesizing CBFs given a system description through a tree-search algorithm inspired by branch-and-cut methods from combinatorial optimization. Finally, we demonstrate our approach on a series of interesting examples of switched affine systems. keywords Control Barrier Functions (CBFs); Switched Systems; Nonsmooth Analysis; Combinatorial Optimization; Safety Enforcement. † † thanks: [ † † address: University of Colorado Boulder, USA. † † address: UC Louvain, Belgium. footnoteinfo]Will be added in subsequent version. , , 1 Introduction We study safety enforcement for switched affine systems through piecewise affine (min-max) control barrier functions. Designing safe controllers is a problem of critical importance for applications such as autonomous vehicles, robotics, and healthcare applications. Safety ensures that system trajectories remain within a predefined safe region, preventing undesired or hazardous outcomes. Control Barrier Functions (CBFs) [ 3 ] form a well-established approach to safety enforcement for autonomous systems. Informally, CBFs are functions over the state-space such that the set of states wherein the CBF is non-positive are control invariant through an appropriate feedback law and, further, the CBF takes on positive values over the unsafe states. In this study, we focus on the synthesis of multiple “min-max” CBFs for switched affine systems, which are formed by taking pointwise minimum/maximum over a set of affine functions. Specifically, we consider switched affine systems of the form \dot{x}(t)=A_{l}x(t)+b_{l} . The control input takes the form of a continuous-time switching signal that selects the mode based on full-state feedback. We use ideas from non-smooth analysis to formulate mathematical conditions for CBFs. We show that given piecewise min/max CBFs, we can extract a state-based feedback in the form of a mapping from states to sets of possible modes that guarantee safety of the closed loop system. We also extend our previous work on synthesis of piecewise affine control Lyapunov functions for switched affine systems to provide an algorithm for synthesizing CBFs from the description of the switched system [ 7 , 19 ] . It is well known that synthesizing polyhedral Lyapunov functions is already a hard problem [ 7 ] . Extending this to polyhedral CBFs introduces additional challenges, as the functions must satisfy more constraints to ensure safety. Our work combines the min- and max- barrier functions to establish a single min-max CBF. Our proposed approach is based on the counterexample-guided inductive synthesis (CEGIS) framework [ 28 ] . This approach is a powerful and widely-used method for synthesizing Lyapunov and barrier functions [ 1 , 2 , 7 , 8 , 11 , 26 ] . Based on the CEGIS approach in our study, the algorithm alternates between a verifier and a learner, iteratively refining candidates until a valid CBF is found or the tree search reaches a predefined expansion limit. Organization: Section 1.1 reviews related work, and Section 1.2 introduces the notations used throughout the paper. Section 2 presents the preliminaries and the problem statement. In Section 3 , we present polyhedral control barrier certificates and pointwise minimum control barriers. The branch-and-bound tree search algorithm is detailed in Section 5 . Section 6 explains the Min-Max multiple barriers. Finally, we evaluate our approach using several examples, including a DC-DC converter, a 2D numerical example, a 3D multi-agent system, and a 6D car velocity system, in Section 7 . Appendices: We have placed the proof of termination of Algorithm 1 (Cf. Section 5 ) in Appendix B . The details of the so-called M-trick for encoding disjunctive constraints as mixed-integer linear programs are in Appendix A . Finally, Appendix C details the CBFs synthesized for our empirical evaluation benchmarks (Cf. Section 7 ). 1.1 Related Work Control Barrier Functions: CBFs were first conceived by Wieland and Allgower [ 33 ] as extensions of the notion of barrier functions [ 25 ] to include control inputs, just as control Lyapunov functions [ 29 , 30 ] extend the classic notion of a Lyapunov function. Ames et al. in [ 4 ] expanded the definition and introduced the notion of safety filters, that have made important contributions towards assured autonomous systems. Since then, the concept has been extended to address high-degree systems [ 34 , 35 ] , uncertain systems [ 31 ] , and systems with actuation constraints [ 3 ] . Notably, many of these works focus on smooth, single CBFs. Nonsmooth Barrier/Lyapunov Functions: Nonsmooth CBFs were first proposed by Glotfelter et al. in [ 17 ] , where the authors introduced the use of min and max operators to define CBFs for multi-agent systems. Subsequent works extended the concept of nonsmooth barrier functions considerably and investigated the ability to form Boolean combinations of CBFs [ 23 , 18 , 16 ] . This paper provides a stronger condition for non-smooth CBFs for switched affine systems and provides algorithms for their synthesis. Sum Of Squares (SOS) Approaches: The use of Sum-Of-Squares approaches for CBF synthesis has been investigaged by many, including the recent work of Clark [ 12 ] . These approaches have studied the problem of synthesizing CBFs for nonlinear systems with polynomial dynamics. However, many existing approaches are restricted to control affine systems and result in bilinear optimization problems that are hard to solve. Here, we focus on an entirely different setup that is characterized by switched affine dynamics and nonsmooth CBFs. Our approach is inspired by a branch-and-bound search with a cutting plane argument to enforce termination. While our approach also suffers from the high complexity of computing CBFs, we synthesize multiple low complexity CBFs and combine them to create CBFs with larger control invariant region. This idea of combining simple barriers to create a larger control invariant set was recently explored by Wajid and Sankaranarayanan for polynomial CBFs using SOS approaches that are very different and complementary to the ideas in this paper [ 32 ] . Neural CBFs: Neural networks have also been utilized for constructing CBFs and CLFs. For example, in [ 1 ] , the authors introduced FOSSIL, a software tool for synthesizing barrier and Lyapunov functions. Their approach employed neural networks during the learner phase of the CEGIS method to find candidate functions. Poonawala et al. in [ 24 ] has utilized single-hidden-layer neural network to synthesize a CLF with ReLU NN for a system with single-layer ReLU NN model. Their approach converts the NN to a PWA function and utilizes nonsmooth analysis to synthesis Lyapunov function and

Read the complete text