On optimal sensor collaboration for distributed estimation with individual power constraints

In the context of distributed estimation, we study the problem of sensor collaboration with individual power constraints, where sensor collaboration refers to the act of sharing measurements with neighboring sensors prior to transmission to a fusion center. In order to find the optimal collaboration strategy consisting of collaboration topology and power allocation scheme, we propose a non-convex formulation in which the estimation distortion is minimized subject to individual power constraints. By exploiting the problem structure, locally optimal collaboration strategies are found via bilinear relaxations and a convex-concave procedure. Numerical examples are provided to show the effectiveness of our approach.