Electrical Engineering-Systems Dept.

סמינר מחלקתי

You are invited to attend a lecture by

Joseph Picard

(Ph.D. student under the supervision of Prof. Anthony J. Weiss, TAU)

on the subject:

Sparsity and Convexity Methods for Localization of Radio Transmitters

In this talk, we address the use of sparsity and convexity principles in the context of geolocation. Several phenomena in geolocation are strongly related to sparsity. For example, in the presence of a few outliers in a set of measurements, the measurement errors build a sparse vector since only the few outliers yield to large entries in this vector. Several other examples also justify the use of sparse models and methods in localization problems.

Compressive sensing, which is not related to geolocation, has recently motivated many research works. In this context, several noticeable advances have been made concerning optimization problems with sparsity constraints. Convex relaxation of sparse problems is a classic tool in compressive sensing and sparse signal representation.

Similarly, convex relaxation is frequently invoked in geolocation. For example, linearizing time-of-arrival measurements is common in range-based localization, since it enables low-complexity estimators. Recent advances in convex optimization justify attempts to perform convex relaxation. Indeed, advanced softwares significantly reduce the computational resources required by convex optimization solvers.

Thus, convex tools are common in localization problems and in sparse problems. Besides, one can also have recourse to sparsity when modeling localization problems. Therefore, it is legitimate to merge these principles and to jointly exploit convexity and sparsity in localization. The algorithms described during this talk rely on both convexity and sparsity principles. We show that their combination is a leading candidate for geolocation with reduced computational costs.