A Two-stage Identification Method for Switched Linear Systems

Zheng Wenju,Ye Hao

2024-07-12

Abstract:In this work, a new two-stage identification method based on dynamic programming and sparsity inducing is proposed for switched linear systems. Our method achieves sparsity inducing in the identification of switched linear systems by the constrained switching mechanism, in contrast to previous optimization-based identification techniques that rely on the rigid data distribution assumption in the parameter space. The proposed mechanism assumes the existence of a minimal interval between adjacent switching instants. First, an efficient iterative dynamic programming approach is used to determine the switching instants and segments using the constrained switching mechanism. Then, each submodel is identified as a combinatorial $\ell_0$ optimization problem, and the true parameter for each submodel is determined by solving the problem. The problem of combinatorial $\ell_0$ optimization is solved by relaxing it into a convex $\ell_1$-norm optimization problem. Furthermore, the unbiasedness of the switched linear system identification is discussed thoroughly with the constrained switching mechanism and a new persistent excitation condition is proposed. Simulation experiments are conducted to indicate that our algorithms exhibit strong robustness against noise.

Systems and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to accurately identify the parameters and switching times of switched linear systems in the presence of noise and when the data distribution assumptions do not hold**.

Specifically, the paper proposes a two - stage identification method based on dynamic programming and sparse optimization to overcome the problems of existing methods being sensitive to noise and relying on strict data distribution assumptions in practical applications. The following is a specific description of the problem:

1. **Challenges in identifying switched linear systems**:
- A switched linear system consists of multiple subsystems, and each subsystem is activated at different time points.
- The difficulty in identifying these systems lies in the fact that it is necessary not only to estimate the parameters of each subsystem but also to determine the switching times and the number of subsystems.
- Existing methods usually rely on strict data distribution assumptions in the parameter space (such as Assumption 1), which are difficult to guarantee in practical applications, especially in the presence of noise.

2. **Solutions proposed in the paper**:
- **First stage**: Use an iterative dynamic programming algorithm to identify the switching times. This method assumes that there is a minimum interval between adjacent switching times (i.e., a constrained switching mechanism), thereby avoiding non - trivial solutions and improving the robustness of identification.
- **Second stage**: Based on the segmented data, use a sparse optimization method (such as ℓ0 - norm optimization) to identify the parameters of each subsystem. To improve computational efficiency and robustness, the ℓ0 - norm optimization problem is relaxed to a convex ℓ1 - norm optimization problem.

3. **Main contributions**:
- Propose a new persistent excitation condition, which is based on the constrained switching mechanism rather than the traditional data distribution assumptions.
- Verify the strong robustness of the proposed method in a noisy environment through simulation experiments.

In summary, this paper aims to solve the key problems in the identification of switched linear systems, that is, how to accurately identify the system parameters and switching times in the presence of noise and when the data distribution assumptions do not hold. By introducing the constrained switching mechanism and the two - stage identification framework, the paper provides a more robust and practical solution.