3D Objects Detection in an Autonomous Car Driving Problem

The autonomous driving problem is one of the most actual problems in both research and industrial fields. Autonomous vehicles are designed to optimize traffic congestion, minimize the number of traffic accidents, provide the ability to transport goods in dangerous conditions for humans, expand the possibilities of using cars for people with disabilities, etc. To successfully solve this problem, it is necessary to consider several subtasks, including localization and mapping, static and dynamic object detection, motion planning, and control. In this paper, we consider one of the tasks of an autonomous vehicle control system: 3D objects detection and localization. We investigate existing detection methods that use heterogeneous information from various sensors such as cameras, depth sensors, and LIDARs. To evaluate the effectiveness of these methods in an autonomous driving problem, we conduct the experimental studies on real and simulated data obtained using CARLA - an open-source simulator for autonomous driving research.