Recursive State Estimation for Lane Detection Using a Fusion of Cooperative and Map Based Data

Abstract

Modern automated and cooperative driver assistance systems (CoDAS) rely deeply on the position estimation. Regardless of absolute positioning accuracy, the relative position in regard to driving environment and other vehicles needs to be of high quality to enable sophisticated functions. Global Navigation Satellite Systems (GNSS) fulfill this demand only partially. In this paper we present an algorithm to accurately infer the driving lane by utilizing Dedicated Short Range Communication (DSRC) and map data alone. We evaluate our approach against simulated and real-life data from Europes largest cooperative vehicle Field Operational Test (FOT): sim TD . This lane detection algorithm will be an enabler for CoDAS functions like collaborative driving and merging developed in the TEAM IP project.

Lateral distance detection model based on convolutional neural network

“…However, before a certain car enters the market, a large number of road tests have to be conducted [9]. To accurately evaluate the performance of the assistant driving systems such as LDWs, LKAs, and ACC, it is necessary to get the precise location of the vehicle between two lane boundaries, namely, a standard benchmark is required [10].…”