Robotics StackExchange | Archived questions

Pure sensor fusion for estimating state

Hi

I am new to ROS and am I was wondering if I can use robot_localization (the ekf) just for fusing two sensor data namely IMU and Wheel odometry and estimate state from that without using any motion model. Also how would this compare to the case where the state is estimated based on an internal motion model and one sensor data.

Thanks

Asked by karthik_ros on 2016-10-09 02:11:08 UTC

Comments

Answers

There's no support for just fusing measurements without a prediction step, no. If you want to limit the effects of the kinematic model, you can fuse absolute pose variables instead of velocities. Your prediction step will be slightly wrong, but then the correction step will pull the state estimate closer to your measurements.

Asked by Tom Moore on 2016-10-12 17:48:14 UTC

Comments