eskf-gps-imu-fusion
eskf-gps-imu-fusion copied to clipboard
fix some formula error
Modifications
- adjust the coordinate system of ins/gps measurement
- add gyro/accel bias compensation into prediction equation
- fix error observation equation and jacobian
Results
@LeatherWang 请问这个mr merge之后,对于输入数据的坐标系是否有要求。例如: accel-0.csv: accel_x (m/s^2),accel_y (m/s^2),accel_z (m/s^2), 是NED还是ENU? gps-0.csv: gps_lat (deg),gps_lon (deg),gps_alt (m),gps_vN (m/s),gps_vE (m/s),gps_vD (m/s),是否依然可以?注意,此处速度为NED gyro-0.csv:gyro_x (deg/s),gyro_y (deg/s),gyro_z (deg/s):相对于IMU body frame的坐标系,or ENU?
@LeatherWang 在理解这个branch的时候遇到了一些困难,不知是否方便请教一下?
@LeatherWang 请问这个mr merge之后,对于输入数据的坐标系是否有要求。例如: accel-0.csv: accel_x (m/s^2),accel_y (m/s^2),accel_z (m/s^2), 是NED还是ENU? gps-0.csv: gps_lat (deg),gps_lon (deg),gps_alt (m),gps_vN (m/s),gps_vE (m/s),gps_vD (m/s),是否依然可以?注意,此处速度为NED gyro-0.csv:gyro_x (deg/s),gyro_y (deg/s),gyro_z (deg/s):相对于IMU body frame的坐标系,or ENU?
根据改动,accel-0 应该是 ENU, 因为imu_flow.cpp
中的TransformCoordinate都被删了。 gyro-0与accel-0一致。 gps-0保持NED, 在tool.h
中对其进行了转换 (注意Eigen::AngleAxisd
复合运算时,旋转轴始终是初始的xyz,因此去掉了.inverse()
)
Thank you for your contribution to this project. I have further refactored the entire algorithm according to your ideas, welcome to review. @LeatherWang