Redundant Multi-Object Detection for Autonomous Vehicles in Structured Environments

This paper presents a redundant multi-object detection method for autonomous driving, exploiting combination of Light Detection and Ranging (LiDAR) stereocamera sensors to detect different obstacles. These are used distinct perception pipelines considering custom hardware/software architecture deployed on self-driving electric racing vehicle. Consequently, the creation local map with respect vehicle position enables development further trajectory planning algorithms. The LiDAR-based algorithm exploits segmentation point clouds ground filtering obstacle detection. stereocamera-based pipeline is based Single Shot Detector using deep learning neural network. presented experimentally validated instrumented during driving maneuvers.