tier4/CalibrationTools

Estimation of EKF Localizer Parameters

yvzksgl opened this issue · 3 comments

yvzksgl commented

Hello,

I have a question about how to estimate ekf_localizer parameters with using deviation_estimator tool. Readme says:
The node also estimates the standard deviation of velocity and yaw rate. This can be used as a parameter in ekf_localizer.
But the node's outputs are imu_corrector.param.yaml and vehicle_velocity_converter.param.yaml files' parameters. Can I transform

velocity_stddev_xx
angular_velocity_stddev_xx
angular_velocity_stddev_yy
angular_velocity_stddev_zz

to

proc_stddev_yaw_c

In short i want to use deviation_estimator tool to optimize ekf_localizer parameters, but i cannot find a way to do that.

knzo25 commented

@kminoda 🙏

  • 👍1
kminoda commented

proc_stddev_yaw_c is a process noise and not observation noise. It is irrelevant to the sensors, but to the kinematics model.
The estimated covariance for IMU and velocity should be reflected in imu_corrector and vehicle_velocity_converter, respectively. Those would be eventually merged into /localization/twist_estimator/twist_with_covariance with the covariance reflected.

So the answer is: Please reflect those estimated parameters not in ekf_localizer but to imu_corrector and vehicle_velocity_converter.

yvzksgl commented

İ understand thank yo for your response. Have a nice week.

  • ❤️1