Tutte seminar - Ting Kei Pong

Convex Relaxation for Sensor Network Localization

Abstract:

In this talk, we consider the problem of sensor network localization, where one locates unknown sensor positions from some known sensor positions (called anchors) and (noisy) distance measurements between neighboring sensors. This problem is NP-hard in general and has received much attention recently. In this talk, we discuss several convex relaxations for the sensor network localization problem. In particular, we compare the strength of SDP, ESDP and sparse-SOS relaxations, and show that zero individual trace certifies accuracy of sensor position when distance measurements are exact. We show by a counterexample that this condition is no longer sufficient in the presence of distance measurement noise, for all three relaxations. We then propose a fix so that small individual trace certifies accuracy of sensor position, when noise is small.

This is based on joint work with Joao Gouveia and Paul Tseng.