Systems and methods for calibration and validation of range sensors and LiDAR sensors mounted on a vehicle that have non-overlapping fields of view (FOV)
Methods and systems calibrating LiDAR sensors mounted on a vehicle and having non-overlapping fields of view (FOVs). The methods include collecting first sensor data corresponding to a calibration environment from a first LiDAR sensor mounted on the vehicle, and second sensor data corresponding to the calibration environment from a second sensor LiDAR sensor mounted on the vehicle. The methods also include transforming the first sensor data to generate a first aligned frame in a global reference frame, and transforming the second sensor data to generate a second aligned frame in the global reference frame. The first reference frame is aligned with the second reference frame to extrinsically calibrate the first LiDAR sensor and the second LiDAR sensor.
1. A method for calibration of LiDAR sensors mounted on a vehicle, the method comprising:
collecting, from a first LiDAR sensor mounted on the vehicle, first sensor data corresponding to a calibration environment;
collecting, from a second sensor LiDAR sensor mounted on the vehicle, second sensor data corresponding to the calibration environment, wherein a first field of view (FOV) of the first LiDAR sensor does not overlap with a second FOV of the second LiDAR sensor;
transforming the first sensor data to generate a first aligned frame in a global reference frame;
transforming the second sensor data to generate a second aligned frame in the global reference frame; and
aligning the first reference frame with the second reference frame to extrinsically calibrate the first LiDAR sensor and the second LiDAR sensor, wherein:
collecting the first sensor data or the second sensor data comprises:
rotating a rotating platform in the calibration environment to a plurality of angular positions, the vehicle being mounted on the rotating platform, and
collecting, using that LiDAR sensor, at each of the plurality of angular positions, a sweep as a collection of LiDAR scans; and
transforming the first sensor data to generate the first aligned frame or transforming the second sensor data to generate the second aligned frame comprises performing inter-sweep alignment between a reference sweep and two or more of a plurality of other sweeps collected by that LiDAR sensor to generate a plurality of aligned subsets of sweeps.
2. The method of claim 1 , further comprising determining the plurality of angular positions based on an FOV of a LiDAR sensor such that two or more sweeps collected by the LiDAR sensor share an overlapping view of the calibration environment.
3. The method of claim 1 , further comprising globally aligning the plurality of aligned subsets of sweeps to generate an aligned frame.
4. The method of claim 3 , wherein globally aligning the plurality of aligned subsets of sweeps comprises performing pose graph optimization to estimate a plurality of LiDAR sensor poses while collecting each of a plurality of sweeps.
5. The method of claim 1 , wherein performing inter-sweep alignment comprises using an iterative closest point technique to generate a transformation between the reference sweep and each of the two or more of the plurality of other sweeps.
6. The method of claim 1 , further comprising determining whether an inter-sweep alignment between the reference sweep and another sweep is successful based on at least one of the following:
an overlap percentage between the reference sweep and the another sweep or a final alignment error.
7. The method of claim 1 , wherein aligning the first reference frame with the second reference frame to extrinsically calibrate the first LiDAR sensor and the second LiDAR sensor comprises using iterative closest point technique to generate a transformation between the first reference frame and the second reference frame.
8. The method of claim 7 , further comprising:
defining an initial transformation based on a model of the calibration environment; and
iteratively refining the initial transformation to generate the transformation.
9. A system for calibration of LiDAR sensors, the system comprising:
a vehicle comprising a first LiDAR sensor and a second LiDAR sensor;
a processor; and
programming instructions stored in a memory and configured to cause the processor to:
collect, from the first LiDAR sensor, first sensor data corresponding to a calibration environment;
collect, from the second sensor LiDAR sensor, second sensor data corresponding to the calibration environment, wherein a first field of view (FOV) of the first LiDAR sensor does not overlap with a second FOV of the second LiDAR sensor;
transform the first sensor data to generate a first aligned frame in a global reference frame;
transform the second sensor data to generate a second aligned frame in the global reference frame; and
align the first reference frame with the second reference frame to extrinsically calibrate the first LiDAR sensor and the second LiDAR sensor by using iterative closest point technique to generate a transformation between the first reference frame and the second reference frame.
10. The system of claim 9 , wherein the instructions to collect the first sensor data or the second sensor data comprise instructions to:
rotate a rotating platform in the calibration environment to a plurality of angular positions, the vehicle being mounted on the rotating platform; and
collect, using that LiDAR sensor, at each of the plurality of angular positions, a sweep as a collection of LiDAR scans.
11. The system of claim 10 , further comprising additional programming instructions that are configured to cause the processor to determine the plurality of angular positions based on an FOV of a LiDAR sensor such that two or more sweeps collected by the LiDAR sensor share an overlapping view of the calibration environment.
12. The system of claim 10 , wherein the instructions to transform the first sensor data to generate the first aligned frame or transform the second sensor data to generate the second aligned frame comprise instructions to perform inter-sweep alignment between a reference sweep and two or more of a plurality of other sweeps collected by that LiDAR sensor to generate a plurality of aligned subsets of sweeps.
13. The system of claim 12 , further comprising additional programming instructions that are configured to cause the processor to globally align the plurality of aligned subsets of sweeps to generate an aligned frame.
14. The system of claim 13 , wherein the instructions to globally align the plurality of aligned subsets of sweeps comprise instructions to perform pose graph optimization to estimate a plurality of LiDAR sensor poses while collecting each of a plurality of sweeps.
15. The system of claim 12 , wherein the instructions to perform inter-sweep alignment comprise instructions to use an iterative closest point technique to generate a transformation between the reference sweep and each of the two or more of the plurality of other sweeps.
16. The system of claim 12 , further comprising additional programming instructions that are configured to cause the processor to determine whether an inter-sweep alignment between the reference sweep and another sweep is successful based on at least one of the following: an overlap percentage between the reference sweep and the another sweep or a final alignment error.
17. The system of claim 9 , further comprising additional programming instructions that are configured to cause the processor to:
define an initial transformation based on a model of the calibration environment; and
iteratively refine the initial transformation to generate the transformation.
18. 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 first LiDAR sensor mounted on a vehicle, first sensor data corresponding to a calibration environment;
collecting, from a second sensor LiDAR sensor mounted on the vehicle, second sensor data corresponding to the calibration environment, wherein a first field of view (FOV) of the first LiDAR sensor does not overlap with a second FOV of the second LiDAR sensor;
transforming the first sensor data to generate a first aligned frame in a global reference frame;
transforming the second sensor data to generate a second aligned frame in the global reference frame; and
aligning the first reference frame with the second reference frame to extrinsically calibrate the first LiDAR sensor and the second LiDAR sensor, aligning the first reference frame with the second reference frame using iterative closest point technique to generate a transformation between the first reference frame and the second reference frame.
19. The computer program product of claim 18 , wherein the instructions that, when executed by a computing device, will further cause the computing device to perform operations comprising:
defining an initial transformation based on a model of the calibration environment; and
iteratively refining the initial transformation to generate the transformation.