On the Predictive Capability of Dynamic Mode Decomposition for Nonlinear Periodic Systems with Focus on Orbital Mechanics

Sriram Narayanan,Mohamed Naveed Gul Mohamed,Indranil Nayak,Suman Chakravorty,Mrinal Kumar

2024-01-25

Abstract:This paper discusses the predictive capability of Dynamic Mode Decomposition (DMD) in the context of orbital mechanics. The focus is specifically on the Hankel variant of DMD which uses a stacked set of time-delayed observations for system identification and subsequent prediction. A theory on the minimum number of time delays required for accurate reconstruction of periodic trajectories of nonlinear systems is presented and corroborated using experimental analysis. In addition, the window size for training and prediction regions, respectively, is presented. The need for a meticulous approach while using DMD is emphasized by drawing comparisons between its performance on two candidate satellites, the ISS and MOLNIYA-3-50.

Systems and Control

What problem does this paper attempt to address?

### Problems the paper attempts to solve

This paper aims to explore the predictive ability of Dynamic Mode Decomposition (DMD) in orbital mechanics, especially for the prediction of nonlinear periodic systems. Specifically, the paper focuses on the Hankel - variant DMD, which identifies the system and makes subsequent predictions by stacking time - delayed observational data. The main objectives of the paper include:

1. **Theoretical analysis**:
- Propose and verify the theory of the minimum number of time delays required for accurately reconstructing the periodic trajectories of nonlinear systems.
- Discuss the selection of the sizes of the training window and the prediction window.

2. **Experimental verification**:
- Verify the theoretical results through experimental analysis.
- Compare the performance of DMD on two candidate satellites (the International Space Station ISS and Molniya - 3 - 50), emphasizing the need for a cautious approach when using DMD.

### Key contributions

1. **Application of DMD in orbital mechanics**:
- Demonstrate how to use DMD to learn from historical trajectory observations and predict the state vector of satellites, providing insights into the predictive ability of DMD in this field.
- Theoretically explain why the periodic trajectories of nonlinear systems can be modeled as linear autoregressive models, and specifically use the DMD algorithm as a data - driven method.

2. **Model parameter selection**:
- Provide a discussion on selecting appropriate DMD model parameters (such as sampling interval, training window size, and prediction window size), especially in the context of Earth - orbit trajectories.

3. **Experimental analysis**:
- Conduct analysis using benchmark dynamic systems (such as the simple pendulum) and model data sets representing various perturbations encountered by candidate satellites.
- Conduct experiments on the data - driven dynamic models generated by DMD for the International Space Station (ISS) and Molniya - 3 - 50, and compare the results of nearly circular orbits (ISS) and highly eccentric orbits (Molniya).

### Mathematical background

The paper discusses in detail how to model the periodic trajectories of nonlinear systems as linear autoregressive (AR) models and use the DMD algorithm for modeling. The main mathematical derivations include:

1. **Fourier series representation of periodic trajectories**:
\[
x(k)=\sum_{m = 0}^{N - 1}a_m e^{j\omega_m k}
\]
where \(a_m\) are Fourier coefficients and \(\omega_m\) are frequencies.

2. **Construction of autoregressive models**:
\[
x(k + l)=\alpha_0x(k)+\alpha_1x(k + 1)+\cdots+\alpha_{l - 1}x(k + l - 1)
\]
Estimate the Fourier coefficients by the least - squares method to determine the parameters of the autoregressive model.

3. **Hankel - DMD algorithm**:
- Construct the Hankel matrix \(H_k\):
\[
H_k=\begin{bmatrix}
X_k\\
X_{k + 1}\\
\vdots\\
X_{k + l - 1}
\end{bmatrix}\in\mathbb{R}^{nl\times s}
\]
- Increase the spatial dimension of the data set through time - delay embedding to improve the spatial resolution.

### Experimental results

The paper uses the normalized 2 - norm error (\(\epsilon\)) as an index to evaluate the training and prediction accuracy. The experimental results show the performance of DMD under different time delays, training window sizes, and prediction window sizes, and are compared with the results of the Fast Fourier Transform (FFT) to confirm that DMD captures the spectral content of the signal.

### Conclusion

Through theoretical analysis and experimental verification, the paper demonstrates the effectiveness of DMD in predicting nonlinear periodic trajectories in orbital mechanics and provides a method for selecting appropriate model parameters. These results have important reference value for satellite operators and orbital dynamics researchers.