Abstract:Sparsity constraints on the control inputs of a linear dynamical system naturally arise in several practical applications such as networked control, computer vision, seismic signal processing, and cyber-physical systems. In this work, we consider the problem of jointly estimating the states and sparse inputs of such systems from low-dimensional (compressive) measurements. Due to the low-dimensional measurements, conventional Kalman filtering and smoothing algorithms fail to accurately estimate the states and inputs. We present a Bayesian approach that exploits the input sparsity to significantly improve estimation accuracy. Sparsity in the input estimates is promoted by using different prior distributions on the input. We investigate two main approaches: regularizer-based MAP, and {Bayesian learning-based estimation}. We also extend the approaches to handle control inputs with common support and analyze the time and memory complexities of the presented algorithms. Finally, using numerical simulations, we show that our algorithms outperform the state-of-the-art methods in terms of accuracy and time/memory complexities, especially in the low-dimensional measurement regime.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper attempts to solve the problem of simultaneously estimating the state and sparse input in Linear Dynamical Systems (LDS) under low - dimensional measurement conditions. Specifically, the paper focuses on how to jointly estimate the state and sparse input of the system from low - dimensional (compressed) measurement data in LDS where the control input is sparse. Due to the existence of low - dimensional measurement data, traditional Kalman filtering and smoothing algorithms cannot accurately estimate the state and input. For this reason, the paper proposes a Bayesian method, which significantly improves the estimation accuracy by using the sparsity of the input. ### Main contributions 1. **Regularization methods**: - **Regularization - based maximum a posteriori (MAP) estimation**: By introducing sparsity priors in the maximum a posteriori estimation of the system state and input, two optimization problems are proposed: ℓ1 regularization and re - weighted ℓ2 regularization. These methods rely on Kalman filtering and smoothing and form low - complexity algorithms. 2. **Bayesian learning methods**: - **Hierarchical Gaussian priors**: Hierarchical Gaussian priors are used to induce sparsity. Two techniques are explored: The first technique uses a Gaussian prior with an unknown variance (as a hyper - parameter), combined with type II maximum likelihood estimation and Kalman smoothing to determine the state and sparse input; the second technique develops a variational Bayes (VB) method, which treats the prior parameters, unknown states and inputs as unobserved variables and infers them through their posterior distributions. 3. **Performance comparison and extension**: - The time and memory complexity of the two methods are analyzed, and numerical simulations show that the Bayesian learning method is superior to the existing state - of - the - art methods in terms of accuracy and time/memory complexity, especially under low - dimensional measurement conditions. - These two methods are further extended to handle sparse control inputs with a common support set, and a similar analysis is carried out. ### Summary By introducing sparsity priors, the paper develops new state and input estimation algorithms that can accurately estimate the state and sparse input of linear dynamic systems under low - dimensional measurement conditions. These methods not only improve the estimation accuracy but also significantly reduce the computational complexity and memory requirements. By combining sparse signal recovery techniques with the Kalman smoothing framework, the paper builds a bridge between sparse signal processing and control theory and promotes the algorithm development in both fields.

Joint State and Sparse Input Estimation in Linear Dynamical Systems

Input and state estimation for linear systems: A least squares estimation approach

The study of joint input and state estimation with Kalman filtering

1 A Kalman Filter Approach for Sparse Input and State Tracking

State Estimation For Nonlinear Systems By Using Dynamic Optimization

Joint State and Input Estimation of Agent Based on Recursive Kalman Filter Given Prior Knowledge

Online Joint State Inference and Learning of Partially Unknown State-Space Models

Input and state estimation for linear systems with a rank-deficient direct feedthrough matrix.

Novel sparseness-inducing dual Kalman filter and its application to tracking time-varying spatially-sparse structural stiffness changes and inputs

Distributed Robust Bayesian Filtering for State Estimation.

Event‐based joint estimation for unknown inputs and states: A distributed recursive filtering method

An information field theory approach to Bayesian state and parameter estimation in dynamical systems

Distributed State Estimation for Sparse Stochastic Systems Based On Compressed Sensing

Input design for Bayesian identification of non-linear state-space models

Sparse Actuator Scheduling for Discrete-Time Linear Dynamical Systems

Joint State Estimation and Noise Identification Based on Variational Optimization

Joint Learning of Linear Time-Invariant Dynamical Systems

Joint State and Parameter Estimation Using the Partial Errors-in-Variables Principle

Joint State and Parameter Estimation of Nonlinear Dynamical Systems Through Alternating Optimization

A Bayesian smoothing for input‐state estimation of structural systems

Kernel Regression-Based State-Space Estimation of Continuous-Time Dynamic Systems From Noisy Sampled State Data