Select thy neighbors: Low complexity link selection for high precision cooperative vehicular localization

Dedicated Short Range Communication (DSRC) based on the IEEE 802.11p standard is expected to stimulate Cooperative Positioning (CP) for emerging cooperative safety

applications. However, the trade-off includes high computational complexity and heavy communication loads while performing exhaustive particle-based information fusion. This paper addresses the problem of selecting the best subset of links to spatial neighbors for "ego" car positioning in a pure GPS-enabled Vehicular Ad hoc NETwork (VANET). Computationally efficient link selection algorithms based on Cram´er-Rao Lower Bound

(CRLB) indicators, including a formulation accounting for neighbors uncertainties, are proposed to cope with various realistic operating contexts. We show that selective fusion significantly reduces the computational complexity and required network traffic with a modest increase in the position error in most cases and an acceptable degradation in the worst-case long-term GPSdenied condition. We also illustrate a typical scenario where evaluating the dispersion of neighboring uncertainties helps to decide the best selection criterion, thus leading to the concept of context-aware selective information fusion.