RadNet - Automotive detection, tracking and prediction using radar data

Project Motivation

Provizio, the partner and sponsor of the project, looks at the self-driving car classification from a different point of view. Instead of automating the process of driving, they saw the need to avoid accidents under all road scenarios as the first step towards autonomous driving. They want to predict and prevent accidents, which helps to avoid 1 million deaths every year. A part of this idea gave birth to this project.   

Project Description

Lidar and Camera, the most used sensors in the car have limitations in adverse weather conditions – rain, snow and fog, and low light conditions. On the other hand, Radar performs better in these conditions and provides long-range detection capability. However, Radar has its own limitations in object classification and resolution due to its Sparse point clouds. This project looks at using Radar and a combination of other automotive sensors for object detection, tracking, and Prediction in urban and motorway scenarios. More specifically, Pedestrians are the most vulnerable road users and trickier to detect and track due to non-linear motion, unlike other road users. Under this project, a key role is given to Pedestrian detection, tracking, and trajectory prediction, which helps to predict and prevent road accidents.    

Status: Ongoing

Team Members

Dr. Ciarán Eising

Dr. Tony Scanlan

Prof. Pepijn van de Ven



Science Foundation Ireland