Towards LiDAR and RADAR Fusion for Object Detection and Multi-object Tracking in CARLA Simulator

Detection and Multi-Object Tracking (DAMOT) systems have a critical role to play in scene understanding in the context of autonomous driving. Modern Autonomous Driving Stacks (ADS) require a software processing unit or module that allows them to understand the data in the environment and convert it into vital information for further decision making. In this context, this work develops a DAMOT module based on Machine Learning techniques, such as DBSCAN or BEV-SORT, that receives information from LiDAR and RADAR sensors in CARLA Simulator. This module uses containerisation techniques with Docker and standard robotics communications with ROS. The performance of the method is evaluated in terms of detection in the AD PerDevKit dataset, developed by the authors.