Closed jack-oquin closed 9 years ago
@shiqizhang6 and I tracked down the bogus odometry readings to https://github.com/segwayrmp/segway_rmp/issues/25. We get bogus odometry readings only when we run the roslaunch command before turning the segway base on. If we turn the segway base on first, it should not happen. A fix for the issue is incoming, but tests should make sure that the segway base on first.
I also talked with @jsinapov a while back, and the localization spread seems a bit much. We'll have to designate someone to test out those parameters soon.
I feel sure that the base was already on before running these tests. If I forget: I habitually kill the original roslaunch; turn on the base; then run everything again.
There are many more checksum errors then usual reported on the segway1 base (flexo). It seems that some bogus messages may be getting through without the checksum algorithm detecting them. That could explain occasional large odometry jumps, observed by @piyushk.
As per William's suggestion in https://github.com/segwayrmp/segway_rmp/issues/26, it's probably best to add an odom filter node to catch these kinds of errors and suppress them.
This problem has not recurred since the failing Segway base was retired.
Closing as a hardware problem.
This was observed recently on flexo, the only Trusty/Indigo system currently running.
To reproduce:
roslaunch bwi_launch segbot_v2.launch
rviz
rviz
After a while, the particle filter spreads out all over the place.
So far, the root cause is unclear. Maybe the navigation stack works differently in Indigo. Maybe the flexo base is periodically generating bogus odometry readings. Maybe it's something else entirely.
Shiqi installed Trusty and Indigo on the clamps laptop. I plan to try that on one of the other segbots.
cc: @shiqizhang6, @mleonetti, @maxsvetlik, @jsinapov