The Linear Quadratic Regulator (LQR) is a classical problem in optimal control theory which deals with operating a linear dynamical system with optimized cost. In this work, we study the discrete-time LQR problem with sparsity constraints on the inputs. This problem has a combinatorial complexity. We develop a convex optimization-based approach to relax the problem into a semidefinite program which can be solved with polynomial complexity. We explore two cases for input sparsity: fixed temporal support and time-varying support. Moreover, we also solve the minimum-energy control problem with sparse inputs. Finally, using numerical simulations, we show that our algorithms give near-optimum performance with very good accuracy and time complexity.

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 maximum a posteriori estimation 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.

We consider the problem of jointly estimating the states and sparse inputs of a linear dynamical system using noisy low-dimensional observations. We exploit the underlying sparsity in the inputs using fictitious sparsity-promoting Gaussian priors with unknown variances (as hyperparameters). We develop two Bayesian learning-based techniques to estimate states and inputs: sparse Bayesian learning and variational Bayesian inference. Through numerical simulations, we illustrate that our algorithms outperform the conventional Kalman filtering based algorithm and other state-of-the-art sparsity-driven algorithms, especially in the low-dimensional measurement regime.