A Multi-modal System Construction for Power Battery Pole Localization

With the rapid development of new energy vehicles, the demand for power batteries is experiencing a surge. However, conventional approaches in power battery production lines, which involve two separate stations for the welding of battery poles and bars, has emerged as a critical bottleneck impeding the enhancement of production efficiency. Flight welding presents a promising solution to this challenge, which, however, requires the equipment to detect precise 3D coordinates of battery poles in real-time when the robotic arm is moving. To remedy this issue, we conduct a multi-modal power battery pole localization system that contains an industrial robotic arm, an industrial camera, and a line laser rangefinder for detecting battery poles accurately. Firstly, we introduce a random sample consensus (RANSAC)-based line laser rangefinder calibration algorithm that adjusts the calculation strategy of standard sphere’s tangential circular radii and then utilize the RANSAC algorithm to effectively filter out unreliable transformation matrices of line laser rangefinder, hence boosting calibration accuracy. Moreover, we propose an interpolation-based timestamp alignment mechanism that synchronizes the detection timestamp of battery poles with the data acquisition timestamp of the line laser rangefinder, mitigating the effects of heterogeneous sensor timestamp discrepancies on the localization precision of poles. Extensive experiments prove that the developed system equipped with the proposed algorithms achieves 0.0247 mm, 0.0333 mm, and 0.0150 mm localization errors for the X, Y, and Z coordinates of battery poles, respectively, at a running speed of 50 mm/s with the robotic arm.