Toward state estimation by high gain differentiators with automatic differentiation

Klaus Röbenack,Daniel Gerbet

DOI: https://doi.org/10.1080/10556788.2024.2320737

2024-03-16

Optimization Methods and Software

Abstract:Most applications of automatic differentiation concern the field of optimization in the broadest sense. This means that many applications only need first and second order derivatives. An exception are control engineering problems, where higher order derivatives are required. This contribution addresses a control engineering problem, namely the estimation of variables that are not measured directly. This problem can be solved with high gain observers and high gain differentiators. They are typically calculated symbolically. We show how automatic differentiation can be used for the implementation of high gain differentiators.

operations research & management science,mathematics, applied,computer science, software engineering

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper primarily addresses the issue of state estimation in control engineering, specifically how to estimate system variables that cannot be directly measured in real-time. The paper proposes a method that utilizes high gain differentiators combined with automatic differentiation techniques to achieve this goal.

### Overview of the Content

1. **Background Introduction**:
- Automatic differentiation techniques are widely used in the field of optimization but usually focus only on first and second-order derivatives.
- Control engineering problems often require higher-order derivatives. This paper explores how to use automatic differentiation techniques to compute higher-order derivatives and apply them to state estimation.

2. **Problem Description**:
- In many practical applications, not all state variables of a system can be directly measured. Therefore, observers need to be designed to estimate these state variables.
- For nonlinear systems, the design of high gain observers requires the computation of Lie derivatives. This paper proposes a new method based on automatic differentiation techniques to compute these Lie derivatives.

3. **Application of Automatic Differentiation Techniques**:
- Using automatic differentiation tools (such as ADOL-C) to compute Lie derivatives, thereby facilitating the design of high gain differentiators.
- This method can avoid the complexity issues associated with traditional symbolic computation.

4. **Design of High Gain Observers and High Gain Differentiators**:
- A high gain observer with a Luenberger structure is designed, which adjusts the observer's state through output error injection.
- Ideally, when the observer gain is sufficiently large, the observer's state will gradually approach the true state of the system.

5. **Numerical Simulation Verification**:
- The effectiveness of the proposed method is verified through numerical simulations of the Duffing system. The experimental results show that, with appropriate parameter selection, the observer can accurately estimate the system's state.

### Summary

This paper proposes a new method that uses automatic differentiation techniques to compute higher-order Lie derivatives, thereby designing high gain differentiators to achieve state estimation for nonlinear systems. This method demonstrates good performance in numerical simulations, especially when dealing with complex nonlinear systems.