Contributions to Unsupervised Online Misalignment Detection and Bumper Error Compensation for Automotive Radar

One of the fundamental sensors utilized in the Advanced Driver Assist System (ADAS) is the radar sensor. Automotive-related functions need highly precise detection and range of traffic and surroundings; otherwise, the whole ADAS performance suffers. The radar placement beneath a bumper or a cover, the age or exposure to accidents or vehicle vibration, vehicle integration, and mounting tolerances will impact the angular performance of the radar sensor. In this research, we present an unsupervised online method for elevation mounting angle error compensation and a method for bumper and environmental error compensation in the azimuth direction. The proposed methods need no specific calibration jig and may be used to replace traditional initial calibration methods; they also enable ongoing calibration throughout the sensor's lifespan. A first proposed standalone method for vertical alignment uses stationary radar targets reflected from the environment to calculate a vertical misalignment angle with a line-fitting algorithm. The vertical mounting error compensation approach delivers two types of correction values: a dynamic value that converges quickly in the case of minor accidents and a more stable correction value that converges slowly but offers a long-term compensation value over the sensor's lifespan. A second proposed solution uses the vehicle velocity and radar targets properties, like relative velocity and measured azimuth angle, to calculate an individual azimuth correction curve. Real-world data collected from drive testing with a 77 GHz series automobile radar was used to analyze the performance of the proposed methods, yielding encouraging results.