Sparsity-promoting sensor management for estimation: An energy balance point of view

In the context of parameter estimation, we study the problem of sensor management under a sparsity-promoting framework, where a sensor being off at a certain time instant is represented by the corresponding column of the estimator coefficient matrix being identically zero. In order to achieve a balance between activating the most informative sensors and uniformly allocating sensor energy, we propose a novel sparsity-promoting approach by adding an ℓ₂-norm penalty function that discourages successive selections of the same group of sensors. We employ the alternating direction method of multipliers (ADMM) to solve the resulting ℓ₂-norm optimization problem, which can then be split into a sequence of analytically solvable subproblems. We finally provide numerical results and comparison with other sensor scheduling algorithms in the literature to demonstrate the effectiveness of our approach.