DriveWorks considers calibration as a representation and estimation of both the (extrinsic) position and orientation as well as the intrinsics parameters for each individual sensor on a vehicle. This is also known as an estimation for vehicle-specific parameters.
A precise knowledge of these values enables accurate mapping of perceived objects in individual sensor frames into other sensor frames, or back into the world frame.
DriveWorks provides multiple APIs, tools, and samples showing how to calibrate sensors and vehicle parameters, and this section of the documentation summarizes existing tools and APIs with the list of supported sensors.
Static Calibration is the process of estimating a sensor's calibration parameters with respect to the vehicle while it is stationary. These parameters include: intrinsics model, extrinsic position and orientation.
Static Calibration consists of multiple steps involving several DriveWorks tools and tutorials to achieve the desired outcome. These for instance include the IMU Calibration Tool to estimate IMU orientation, or the Camera Calibration Tools to estimate camera intrinsics parameters, position and orientation. The Static Camera Calibration Tutorial tutorial provides assistance in setting up an environment optimized for calibration.
Dynamic Calibration, also known as self-calibration, is the process of estimating a sensor's calibration parameters with respect to the vehicle while it is non-stationary. This process does not require a specifically prepared environment to optimize calibration. Instead, certain driving maneuvers can influence how these dynamic calibration estimates are obtained. These maneuvers can include extensive turns, stronger acceleration, or smooth driving in a straight line. Although DriveWorks APIs and tools do not require you to perform these maneuvers, doing so may result in quicker calibration results.
The following is a list of sensors supported by the static calibration and dynamic calibration DriveWorks APIs and tools.
DriveWorks provides a collection of APIs and tools for both dynamic and static IMU calibration. In general, any IMU sensor supported by the Sensor Abstraction Layer (SAL) is considered to be supported.
For more information regarding static IMU Calibration, please refer to IMU Calibration Tool.
IMU Self-Calibration covers in great detail how IMU self-calibration is implemented and which driving maneuvers are beneficial for acceptance of an accurate estimate.
Following is a list of sensors and DoFs that can be calibrated for IMU sensors using the APIs as provided by the calibration engine, see dwCalibrationEngine_initializeIMU.
These IMU sensors were tested and supported by both the static calibration tool as well as the self-calibration API:
| Sensor | Calibration support | DW version |
|---|---|---|
| XSens MTi-G-710 | roll,pitch,yaw | >= 2.0 |
| Continental SC13 | roll,pitch,yaw | >= 3.0 |
DriveWorks provides support for the camera modules as listed in dwCameraType. It also provides support for a custom camera through external drivers. Calibration tools or APIs do not differentiate for specific camera types, however they do depend on the availability of frames captured with these cameras. They hence indirectly have a dependency on a working Sensor Abstraction Layer (SAL).
DriveWorks supports multiple mathematical models of the geometrical relationship between pixels and optical rays, also known as intrinsic camera models, see Camera Model.
DriveWorks's static calibration tools support the f-theta (dwFThetaCameraConfig) and pinhole (dwPinholeCameraConfig) camera models. Estimated parameters of these models are also known as intrinsic calibration parameters.
To additionally support mapping objects from camera space to the rig coordinate system, all camera position and orientation parameters on the vehicle have to be calibrated as well. These parameters are also known as the camera's extrinsic calibration parameters.
Please refer to Camera Calibration Tools for the list of the tools provided for static camera calibration. The step-by-step guide Static Camera Calibration Tutorial provides guidelines and best practices on how to successfully perform intrinsic as well as extrinsic camera calibration. Static camera calibration can be performed on any camera which is supported by the Sensor Abstraction Layer (SAL).
Self-calibration is a method of estimating camera calibration parameters while the vehicle performs regular driving maneuvers.
DriveWorks provides APIs, see Calibration Interface, to implement self-calibration in custom applications.
A collection of samples, for example Camera Calibration Sample, demonstrates how the DriveWorks API can be used to calibrate recorded sessions. An explanation of how self-calibration is implemented, as well as some code-snippets, can be found at Feature-based Camera Self-Calibration.
Following is a list of supported and verified sensors (cameras absent from the table below, but which are listed in dwCameraType are not fully verified at this time):
| Sensor | FOV | Orientation | Dynamic calibration | DW version |
|---|---|---|---|---|
| AR0231 | 60deg | front/rear | roll,pitch,yaw,height | >= 2.0 |
| AR0231 | 60deg | side/cross | roll,pitch,yaw | >= 3.0* |
| AR0231 | 120deg | front/rear | roll,pitch,yaw,height | >= 2.0 |
| AR0231 | 120deg | side/cross | roll,pitch,yaw | >= 3.0* |
| AR0820 | 30deg | front/rear | pitch,yaw | >= 3.0 |
| AR0820 | 30deg | side/cross | n.a. | n.a. |
| AR0820 | 70deg | front/rear | roll,pitch,yaw,height | >= 3.0 |
| AR0820 | 70deg | side/cross | roll,pitch,yaw | >= 3.0* |
| AR0820 | 120deg | front/rear | roll,pitch,yaw,height | >= 3.0 |
| AR0820 | 120deg | side/cross | roll,pitch,yaw | >= 3.0* |
| IMX390 | 190deg | front/rear | roll,pitch,yaw,height | >= 3.2** |
| IMX390 | 190deg | side/cross | roll,pitch,yaw | >= 3.2** |
Notes:
yaw in the ranges of (-30,30) and (150,-150) degrees.yaw is in the ranges of (120,60) and (-120,-60) degrees.yaw is in the ranges of (60,30), (-60,-30), (120,150) and (-120,-150) degrees.DriveWorks provides support for a variety of lidar sensors through it's Sensor Abstraction Layer (SAL) interface, see Lidar Sensor.
There are no tools provided to calibrate lidars statically yet.
APIs to perform dynamic lidar calibration can be found at Calibration Interface. See Lidar Self-Calibration for a detailed explanation of the algorithms to self-calibrate, as well as verification methodology of lidar dynamic calibration.
Following is a list of supported and verified lidars for self-calibration:
| Sensor | Calibration support | DW version |
|---|---|---|
| Velodyne HD32 | roll,pitch,yaw,height | >= 2.0 |
| Velodyne HD64 | roll,pitch,yaw,height | >= 2.0 |
| Velodyne VLS128 | roll,pitch,yaw,height | >= 2.0 |
| Ouster OS2-128 | roll,pitch,yaw,height | >= 3.2* |
| Luminar H3 | roll,pitch,yaw,height | >= 3.2* |
Notes:
There are no tools provided to calibrate radars statically yet.
DriveWorks provides APIs to perform radar self-calibration. See Radar Self-Calibration for a more detailed explanation of how self-calibration of radars is implemented and verified.
Following is a list of supported and tested radars for self-calibration:
| Sensor | Calibration support | DW version |
|---|---|---|
| ARS430 (CAN) | yaw, wheel-radii | >= 2.0 |