Computer-assisted proofs for Lyapunov stability via Sums of Squares certificates and Constructive Analysis

Grigory Devadze,Victor Magron,Stefan Streif

2024-08-01

Abstract:We provide a computer-assisted approach to ensure that a given continuous or discrete-time polynomial system is (asymptotically) stable. Our framework relies on constructive analysis together with formally certified sums of squares Lyapunov functions. The crucial steps are formalized within of the proof assistant Minlog. We illustrate our approach with various examples issued from the control system literature.

Optimization and Control,Logic in Computer Science

What problem does this paper attempt to address?

The paper attempts to address the problem of how to prove the (asymptotic) stability of a given discrete-time polynomial system using computer-aided methods. Specifically, the authors propose a method based on constructive analysis and formal verification, utilizing Sums of Squares (SOS) certificates to construct Lyapunov functions and verify their correctness within the Minlog proof assistant.

### Main Issues
1. **System Stability Verification**: How to ensure that a given discrete-time polynomial system is stable?
2. **Construction of Lyapunov Functions**: How to construct Lyapunov functions using SOS techniques?
3. **Formal Verification**: How to verify the correctness of the constructed Lyapunov functions in the Minlog proof assistant?

### Solutions
1. **SOS Techniques**: Construct Lyapunov functions using SOS techniques to ensure they meet stability conditions.
2. **Formal Verification Framework**: Implement a formal verification framework in the Minlog proof assistant to ensure the correctness of the constructed Lyapunov functions.
3. **Algorithm Implementation**: Provide an algorithm named `exactLyapunov` to compute exact weighted rational SOS Lyapunov functions.

### Specific Steps
1. **Construction of Lyapunov Functions**: Construct Lyapunov functions \( V \) using SOS techniques, ensuring they are positive definite and \( V - V \circ f \) is non-negative.
2. **Formal Verification**: Verify the correctness of the constructed Lyapunov functions in Minlog, ensuring they meet all necessary conditions.
3. **Algorithm Implementation**: Implement the `exactLyapunov` algorithm to ensure it can compute exact weighted rational SOS Lyapunov functions.

### Related Work
- **Formalized Dynamic Systems**: Work by other researchers on formalizing dynamic systems, such as using tools like Frama-C, Isabelle/HOL for verifying the correctness of numerical solutions.
- **Formalization of Constructive Analysis**: Formalization of constructive real analysis in proof assistants like Coq and Minlog.
- **Certification of Polynomial Systems**: Using SOS techniques and semidefinite programming relaxation methods to verify the stability of polynomial systems.

### Contributions
- **Algorithm Development**: Propose and implement the `exactLyapunov` algorithm for computing exact weighted rational SOS Lyapunov functions.
- **Formal Verification Framework**: Develop a formal verification framework in Minlog for verifying the stability of discrete-time polynomial systems.
- **Practical Application**: Demonstrate the effectiveness and practicality of the method through specific examples.

In summary, the paper provides a reliable computer-aided method to prove the stability of discrete-time polynomial systems by combining SOS techniques and formal verification methods.