robot_localization as simple filter without fusion

Question

Can robot_localization be used without fusing multiple sources, ie, as a simple filter for odometry alone? I do not need or want to fuse an IMU with odometry, rather I simply want to smooth out the existing odometry.

Although the use-case may seem odd, I have a motion-capture system that gives more-or-less a real-time transform for the robot. However the mocap reported transform jitters a bit, and rarely it will jump in and out of an outlier. I could address this with a moving-average filter (too slow) or another more traditional filter (some might argue Kalman may be overkill) but I'd prefer to use the robot_localization infrastructure. In the future I may have other sensors to fuse and it makes sense to architect this system around RL.

---

Originally posted by codenotes on ROS Answers with karma: 261 on 2015-12-27

Post score: 2

---

Original comments

Comment by Devin on 2016-02-29:
Could you show me your configuration files? I have the same question. But without IMU, it doesn't work.Thanks.

Answer 1

Yes, you can do that. Just make sure the covariance on the measurements is high enough to allow the smoothing to happen. If it's too small, the filter output will also be jittery.

---

Originally posted by Tom Moore with karma: 13689 on 2016-02-05

This answer was ACCEPTED on the original site

Post score: 2

---

Original comments

Comment by codenotes on 2016-02-06:
Thanks...very good to know. So simply by omitting the IMU in the configuration, it will do that? Great...because sometimes you just need a filter. I guess is just uses the measurement itself as the model.

Comment by Tom Moore on 2016-02-06:
Yep, there are no restrictions on the number of sensor inputs. It uses its own omnidirectional kinematic model, which is independent of the measurement sources. It predicts with the model, then corrects with your sensor data.