Radar Sensing for Intelligent Vehicles in Urban Environments

Area of Science:

• Robotics and Autonomous Systems
• Sensor Fusion
• Computer Vision

Background:

• Traditional sensors like LiDAR, stereovision, and sonar face limitations in adverse environmental conditions such as dust, fog, heavy snow, and low light.
• Reliable perception is critical for the safe operation of Advanced Driving Assistance Systems (ADAS) and autonomous vehicles (AVs).

Purpose of the Study:

• To develop and validate a novel ground and obstacle segmentation method utilizing radar sensing.
• To overcome the environmental and operational limitations of existing perception systems.
• To enable robust vehicle perception without requiring external navigation or precise sensor calibration.

Main Methods:

• A new algorithm for ground and obstacle segmentation is proposed, operating directly within the radar sensor's native frame.
• The method avoids the need for synchronized navigation data or detailed vehicle-to-radar calibration parameters.
• It also removes geometric constraints imposed by previous state-of-the-art techniques.

Main Results:

• The proposed radar-based segmentation method demonstrated effective performance in diverse urban driving scenarios.
• Experimental validation confirmed the algorithm's ability to segment ground and obstacles accurately.
• The approach proved robust even in challenging environmental conditions where other sensors might fail.

Conclusions:

• The novel radar sensing approach offers a significant advancement for vehicle perception systems.
• It provides a viable and robust solution for ground and obstacle segmentation, particularly in adverse weather conditions.
• The method holds strong potential for enhancing the safety and reliability of ADAS and AVs.