Addressing errors in automated sensor data for real-time traffic state estimation using dynamical systems approach

Developments in the area of intelligent transportation system in India are constrained due to the difficulty in accurate automated data collection. Many existing automated sensors may not be accurate under Indian traffic conditions due to their heterogeneity and less lane discipline, resulting in erroneous data. Thus, there is a need to develop tools and estimation schemes that can address these errors and still be able to generate reasonably accurate results. The present study addresses this issue, considering the real-time estimation of speed and density. A dynamical systems approach using the Kalman filtering technique was developed. The implementation was done using data fusion where location-based and spatial traffic variables were used. The estimated values were compared with field data and it was observed that the proposed method was reasonably accurate in the presence of erroneous data.