System and method for ground and free-space detection
Improved processing of sensor data (e.g., LiDAR data) can be used to distinguish between free space and objects/hazards. Autonomous vehicles can use such information for performing autonomous driving and/or parking operations. LiDAR data can include a plurality of range measurements (e.g., forming a 3D point cloud). Each of the range measurements can correspond to a respective LiDAR channel and azimuth angle. The processing of LiDAR data can include identifying one or more paths as candidate ground paths based on one or more path criteria. The processing of LiDAR data can also include identifying one or more of the plurality of range measurements as ground points or non-ground points based on the one or more paths identified as candidate ground paths and based on one or more point criteria.
1. A system configured to be incorporated into a vehicle, the system comprising:
one or more light detection and ranging (LiDAR) sensors comprising one or more LiDAR channels; and
one or more processors coupled to the one or more LiDAR sensors, the processor configured to process ranging data from the one or more LiDAR sensors, the ranging data including a plurality of paths, each respective path comprising a range measurement for each of the of the one or more LiDAR channels at an azimuth angle of the respective path, wherein processing the ranging data comprises the steps of:
selecting a first subset of paths of the plurality of paths to include a pair of non-adjacent paths separated by a distance;
selecting a second subset of paths of the plurality of paths;
for each respective path of the first subset of paths:
decomposing the ranging data of the path into height projection data and ground projection data,
identifying the path as a candidate ground path based on one or more path criteria by computing derivatives along the path using differentials of the path's height projection data and differentials of the path's ground projection data between adjacent LiDAR channels, and
in the event the path is identified as a candidate ground path, identifying one or more of the range measurements of the path as ground points or non-ground points based on one or more point criteria; and
for each respective path of the second subset of paths:
foregoing decomposing the ranging data of the path into height projection data and ground projection data,
foregoing identifying the path as either a candidate ground path or a non-candidate ground path, and
foregoing identifying the one or more range measurements of the path as ground points or non-ground points;
in the event that the pair of non-adjacent paths are identified as ground points, selecting the first subset of paths to include one or more intervening paths between the pair of non-adjacent path;
in the event that the pair of non-adjacent paths are identified as non-ground points, selecting the second subset of paths to include the one or more intervening paths between the pair of non-adjacent paths.
2. The system of claim 1 , wherein the one or more processors are further configured to:
generate navigation commands for the vehicle in accordance with the identification of a subset of the ranging data as ground points or non-ground points.
3. The system of claim 1 , wherein the one or more path criteria includes a criterion that is satisfied when the sum of the derivatives is less than a threshold, wherein identifying the path as a candidate ground path comprises
identifying the respective path as a candidate ground path when the sum of derivatives along the respective path is less than the threshold.
4. The system of claim 1 , wherein identifying the range measurements of the path as ground points or non-ground points comprises:
for a range measurement of a respective LiDAR channel of the one or more LiDAR channels of the path:
computing differentials of the ground projection data between adjacent azimuth angles for the respective LiDAR channel; and
computing a plurality of derivatives along the respective LiDAR channel using the differentials of the ground projection data and a differential between adjacent azimuth angles.
5. The system of claim 4 , wherein the one or more point criteria includes a criterion that is satisfied when a first computed derivative of the plurality of derivatives and a second computed derivative of the plurality of derivatives along the respective LiDAR channel are each less than a threshold, wherein identifying the range measurement as a ground point or a non-ground point comprises:
identifying the range measurement as a ground point when the first computed derivative and the second computed derivative are each less than the threshold; and
identifying the range measurement as a non-ground point when the first computed derivative or the second computed derivative is at or greater than the threshold.
6. The system of claim 1 , wherein the one or more processors are further configured to:
identify free space based on the minimum distance along the one or more paths to one or more non-ground points.
7. A method comprising:
processing ranging data from one or more light detection and ranging (LiDAR) sensors, the LiDAR sensors comprising one or more LiDAR channels, the ranging data including a plurality of paths, each respective path comprising a range measurement for each of the of the one or more LiDAR channels at an azimuth angle of the respective path, wherein processing the ranging data comprises the steps of:
selecting a first subset of paths of the plurality of paths to include a pair of non-adjacent paths separated by a distance;
selecting a second subset of paths of the plurality of paths;
for each respective path of the first subset of paths:
decomposing the ranging data of the path into height projection data and ground projection data,
identifying the path as a candidate ground path based on one or more path criteria by computing derivatives along the path using differentials of the path's height projection data and differentials of the path's ground projection data between adjacent LiDAR channels, and
in the event the path is identified as a candidate ground path, identifying one or more of the range measurements of the path as ground points or non-ground points based on one or more point criteria; and
for each respective path of the second subset of paths:
foregoing decomposing the ranging data of the path into height projection data and ground projection data,
foregoing identifying the path as either a candidate ground path or a non-candidate ground path, and
foregoing identifying the one or more range measurements of the path as ground points or non-ground points;
in the event that the pair of non-adjacent paths are identified as ground points, selecting the first subset of paths to include one or more intervening paths between the pair of non-adjacent paths; and
in the event that the pair of non-adjacent paths are identified as non-ground points, selecting the second subset of paths to include the one or more intervening paths between the pair of non-adjacent paths.
8. The method of claim 7 , further comprising:
generating navigation commands for a vehicle in accordance with the identification of a subset of the ranging data as ground points or non-ground points.
9. The method of claim 7 , wherein the one or more path criteria includes a criterion that is satisfied when the sum of the derivatives is less than a threshold, wherein identifying the path as a candidate ground path comprises
identifying the respective path as a candidate ground path when the sum of derivatives along the respective path is less than the threshold.
10. The method of claim 7 , identifying the range measurements of the path as ground points or non-ground points comprises:
for a range measurement of a respective LiDAR channel of the one or more LiDAR channels of the path:
computing differentials of the ground projection data between adjacent azimuth angles for the respective LiDAR channel; and
computing a plurality of derivatives along the respective LiDAR channel using the differentials of the ground projection data and a differential between adjacent azimuth angles.
11. The method of claim 10 , wherein the one or more point criteria includes a criterion that is satisfied when a first computed derivative of the plurality of derivatives and a second computed derivative of the plurality of derivatives along the respective LiDAR channel are each less than a threshold, wherein identifying the range measurement as a ground point or a non-ground point comprises:
identifying the range measurement as a ground point when the first computed derivative and the second computed derivative are each less than the threshold; and
identifying the range measurement as a non-ground point when the first computed derivative or the second computed derivative is at or greater than the threshold.
12. The method of claim 7 , the method further comprising:
identifying free space based on the minimum distance along the one or more paths to one or more non-ground points.
13. A non-transitory computer readable storage medium storing instructions, which when executed by one or more processors, cause the one or more processors to perform a method, the method comprising:
processing ranging data from one or more light detection and ranging (LiDAR) sensors, the LiDAR sensors comprising one or more LiDAR channels, the ranging data including a plurality of paths, each respective path comprising a range measurement for each of the of the one or more LiDAR channels at an azimuth angle of the respective path, wherein processing the ranging data comprises the steps of:
selecting a first subset of paths of the plurality of paths to include a pair of non-adjacent paths separated by a distance;
selecting a second subset of paths of the plurality of paths;
for each respective path of the first subset of paths:
decomposing the ranging data of the path into height projection data and ground projection data,
identifying the path as a candidate ground path based on one or more path criteria by computing derivatives along the path using differentials of the path's height projection data and differentials of the path's ground projection data between adjacent LiDAR channels, and
in the event the path is identified as a candidate ground path, identifying one or more of the range measurements of the path as ground points or non-ground points based on one or more point criteria; and
for each respective path of the second subset of paths:
foregoing decomposing the ranging data of the path into height projection data and ground projection data,
foregoing identifying the path as either a candidate ground path or a non-candidate ground path, and
foregoing identifying the one or more range measurements of the path as ground points or non-ground points;
in the event that the pair of non-adjacent paths are identified as ground points, selecting the first subset of paths to include one or more intervening paths between the pair of non-adjacent paths; and
in the event that the pair of non-adjacent paths are identified as non-ground points, selecting the second subset of paths to include the one or more intervening paths between the pair of non-adjacent paths.