An Improved D-S Based Vehicular Multi-Sensors' Perceptual Data Fusion for Automated Driving Decision-Making

In the automated driving scenario, to cope with the complex road condition and obtain the comprehensive and accurate identified result to make the automated driving decision, how to fuse the heterogeneous perceptual date perceived by vehicular multi-sensors efficiently and make automated driving decision reliably is the main issue. In this paper, a mathematical model is proposed to abstract the heterogeneous perceptual data from different vehicular sensors into a series of vectors (data set), which have a uniform measurement accuracy, range, and output form and map them to a corresponding automated driving decision. then we fuse abstracted perceptual data of vehicular multi-sensors and make the automated driving decision based on a D-S (Dempster-Shafer) evidence theory. However, using the classical D-S evidence theory to fuse may exist large fusion computational resource consumption and weak ability to fuse high-conflict perceptual data, it causes fusion latency to increase and accuracy of decision-making to decrease. So an improved D-S evidence theory is proposed to overcome the problem above for satisfying the requirements for real-time and reliability in the automated driving scenario.