Floating car and camera data fusion for non-parametric route travel time estimation

The paper proposes a non-parametric route travel time estimation method based on fusion of floating car data (FCD) and automated number plate recognition (ANPR) data. Today´s traffic management utilizes heterogeneous data collection systems which can be stationary or mobile. Each data collection system has its own advantages and disadvantages. Stationary sensors usually have less measurement noise than mobile sensors but their network coverage is limited. On the other hand, mobile sensors, commonly installed in fleet vehicles, cover relatively wider areas of the network but they suffer from low penetration rate and low sampling frequency. Traffic state estimations can benefit from fusion of data collected by various sources as they complement each other. The proposed estimation method is implemented using FCD from taxis and the ANPR data from Stockholm, Sweden. The results suggest that the fusion increases the robustness of the estimation, meaning that the fused estimates are always better than the worst of the two (FCD or ANPR), and it sometimes outperforms the two single sources.