mmSCALE: Self-Calibration of mmWave Radar Networks from Human Movement Trajectories

We present mmSCALE, a practical self-calibration method that automatically estimates the relative position and orientation of a network of millimeter wave (mmWave) radars by post-processing the trajectories of detected targets that move within the radars’ fields of view (FoVs). This is a key component of multi-device mmWave radar deployments for indoor human sensing. As commercial mmWave radars have limited range (up to 6–8 m) and are subject to occlusion, covering large indoor spaces requires multiple radars. A fully self-contained system should estimate the location and orientation of each radar with no intervention by a human operator. To solve this problem, mmSCALE fuses target detections from multiple radars, yielding median errors of 0.18 m and 2.86° for radar location and orientation estimates, respectively. For this, mmSCALE requires no specific target trajectories or controlled conditions, it autonomously assesses the calibration quality over time, and is robust to occlusion and to the presence of multiple subjects.