SENSOR FUSION FOR DYNAMIC OBJECT DETECTION IN AUTONOMOUS DRIVING

Abstract

Autonomous Driving has risen in popularity over the years with the rise of computing capabilities in embedded system and sensor availability as well. The required tasks for accomplishing autonomous driving include perception, planning and controls. This thesis focuses on the perception task, specifically sensor fusion algorithm for object detection of dynamic obstacles (using multi-modal sensors) for real-world implementation and compares the performance between a typical autonomous vehicle and an untypical vehicle - an autonomous bus. The author's key contributions includes helping to develop a dataset from an autonomous bus and modifications of current sensor fusion algorithms in 4 ways. The outcome of the project is also shown in the publications and videos produced found in the appendices.