Author(s)

Zhe Xu and Xuhui Ran

Corresponding Author

Zhe Xu

ABSTRACT

Aiming at the problem that four-layer LiDAR is difficult to identify neighboring obstacles in the obstacle recognition of complex road environment, an obstacle recognition algorithm considering both adaptive and real-time characteristics is proposed. The algorithm clusters obstacles detected based on Dempster-Shafer Theory (DST) and conflict coefficients in a grid map through segment based connected region labeling algorithm, which reduces the number of neighborhood searches and increases the real-time performance of the algorithm. The Borges distance threshold is improved by combining the characteristics of obstacle data, and adaptive recognition is used to optimize the recognition of neighboring obstacles through the improved distance threshold and the distribution characteristics of obstacle data. The experimental results show that the algorithm can significantly improve the recognition of neighboring obstacles and the real-time performance of the algorithm meets the real-time requirements of detecting obstacles when unmanned vehicles are driving.

KEYWORDS

Driverless vehicle; LiDAR; environmental awareness; Dempster-Shafer Theory; grid map