Team channel-SLAM: A cooperative mapping approach to vehicle localization

Vehicle positioning is considered a key element in autonomous driving systems. While conventional positioning requires the use of GPS and/or beacon signals from network infrastructure for triangulation, they are sensitive to multipath and signal obstruction. However, recent proposals like the Channel-SLAM method showed it was possible in principle to in fact leverage multi-path to improve positioning of a single vehicle. In this paper, we derive a cooperative Channel-SLAM framework, which is referred as Team Channel-SLAM. Different from the previous work, Team Channel-SLAM not only exploits the stationary nature of reflecting objects around the receiver to characterize the location of a single vehicle through multi-path signals, but also capitalizes on the multi-vehicle aspects of road traffic to further improve positioning. Specifically, Team ChannelSLAM exploits the correlation between reflectors around multiple neighboring vehicles to achieve high precision multiple vehicle positioning. Our method uses affinity propagation clustering and cooperative particle filter. The new framework is shown to give substantial improvement over the single vehicle positioning situation.