Fast Dynamic Calibration of Millimeter-Wave Radar External Parameters: Research and Applications

With the rapid development of intelligent connected technologies, the application of millimeter-wave radar in roadside perception systems has become increasingly widespread [1], [2]. Known for its excellent anti-interference capabilities and resistance to adverse weather conditions, millimeter-wave radar has become a key sensor in these systems [3]. However, the accuracy of radar data is still limited by the calibration of the coordinate system, especially in dynamic environments [4]. Traditional calibration methods rely on manual adjustments, which are inefficient [5]. This paper proposes an automated dynamic millimeter-wave radar calibration method based on deep reinforcement learning, combined with multi-sensor data fusion, to address the high-precision calibration issue between the radar coordinate system and the world coordinate system in dynamic environments. Experiments on open roads show that both the new method and the method incorporating deep reinforcement learning (DQN) significantly improve calibration performance. The calibration accuracy increased by 11.11% and 48.15%, respectively, compared to traditional methods, while the calibration speed was also greatly improved. The proposed method effectively reduces human intervention, lowers system deployment costs, and demonstrates strong scalability, making it highly valuable for practical applications in intelligent connected vehicles and roadside perception systems.