I have a robot with a Honeycomb lidar. The mirror inside the lidar drifts as the lidar works for longer time, and causes point clouds from different beam sides misaligned. If the lidar is well calibrated, point clouds from different beam sides should be visually aligned. I want to design an alert system that can send out an alert if point clouds from different beam sides are visually misaligned. I can apply noise removal and registration to point clouds and find out the transformation between point clouds. I can also have the system send out the alert based on the transformation. The problem is that how bad the transformation should be to send out the alert. The criteria for transformation is different for different environments even if we remove noise from point cloud.