Bayesian Network based speed estimation in case of C2X data at very low equipment rates

The estimation of the traffic state, particularly speed and vehicle counts, is important for modern traffic light control (TLC) algorithms. Recent developments in the field of so called self-organizing TLC focus on using cooperative traffic data transmitted via Carto-Infrastructure (C2I) or Car-to-X (C2X) communication. In this context, vehicles periodically exchange their GPS position and speed via so called Cooperative Awareness Messages (CAM) in the EU (a.k.a. Basic Safety Messages (BMS) in the US) transmitted on a vehicular-specific extension to the WiFi ad-hoc mode IEEE802.11p1. Equipped with such technology, TLC may intercept CAM/BSM and use the contained data to estimate its local traffic state. This technology is yet emerging and car manufacturers are not expecting a sufficient penetration at the time future self-organizing TLC systems will be deployed. In this paper, a novel approach proposes to complement missing C2X technology with Bluetooth technology (BT), which is already widely available in all modern vehicles as well as smartphones. BT data has notably been tested for traffic data acquisition for years, but neither provide the same type nor the same quality of C2X cooperative data. Accordingly, we propose to rely on a stochastic data fusion engine based on Bayesian Networks (BN) to enrich and integrate traffic state data from a low C2X equipment rate (about 1%) and moderate BT equipment rate (30%). One of the challenges addressed is to be able to extrapolate mobility from detection processes, and then extract the estimated BT and C2X speed likelihoods. Using BN, we will propose to link such likelihood to different a-priori knowledge of mobility states. Another challenge is to integrate the BT and the IEEE802.11p is known as ITS-G5 in the EU and as DSRC in the US.