Systems and methods for calibration and validation of non-overlapping range sensors of an autonomous vehicle

A system for validating calibration of a LiDAR sensor, the system comprising:

• a LiDAR sensor mounted on a vehicle;
• at least one processor;
• and
• programming instructions stored in a memory and configured to cause the processor to:
• collect, from the LiDAR sensor, sensor data corresponding to a calibration environment, wherein the sensor data is collected over a plurality of point cloud sweeps scanned by the LIDAR sensor capturing different areas of the calibration environment, wherein each of the plurality of point clouds sweeps including a plurality of points and is associated with a collection time stamp;
• generate, from the sensor data, a reconstructed point cloud corresponding to the calibration environment;
• segment the reconstructed point cloud of the calibration environment to generate a plurality of planes;
• determine, for each of the plurality of point cloud sweeps, a normal distance as an average of point-to-plane distances measured between each point of the point cloud sweep and one or more of the plurality of planes;
• determine, as a validation score, an average of using a plurality of the normal distances determined for the plurality of point cloud sweeps;
• and
• validate a calibration of the LiDAR sensor in response to determining that the validation score is less than a threshold.

A computer program product comprising a non-transitory computer-readable medium that stores instructions that, when executed by a computing device, will cause the computing device to perform operations comprising:

• collecting, from a LiDAR sensor, sensor data corresponding to a calibration environment, the sensor data being collected over a plurality of point cloud sweeps scanned by the LIDAR sensor capturing different areas of the calibration environment, wherein each of the plurality of point clouds sweeps including a plurality of points and is associated with a collection time stamp;
• generating, from the sensor data, a reconstructed point cloud corresponding to the calibration environment,
• segmenting the reconstructed point cloud of the calibration environment to generate a plurality of planes;
• determining, for each of the plurality of point cloud sweeps, a normal distance as an average of point-to-plane distances measured between each point of the point cloud sweep and one or more of the plurality of planes;
• determining, as a validation score, an average of a plurality of the normal distances determined for the plurality of point cloud sweeps;
• and
• validating a calibration of the LiDAR sensor in response to determining that the validation score is less than a threshold.