Wrist calibration controller shouldn't crash realtime (ros ticket #3558)

PR2 / pr2_controllers

The controllers that run in realtime on the PR2 and supporting packages.

47 stars 35 forks source link

Wrist calibration controller shouldn't crash realtime (ros ticket #3558) #265

Closed ahendrix closed 11 years ago

ahendrix commented 11 years ago

Bringing up on PRE, I got the following error:

'''[FATAL] 1263442663.234200000: k = 4729.3333 is outside of [0,1]'''

It looks like this error is being generated from:

'''pr2_calibration_controllers/src/pr2_calibration_controllers.cpp:277'''

This controller should probably report the error to diagnostics or somewhere. Crashing realtime in not ok.

trac data:

Owner: sglaser
Reporter: vpradeep
Reported at: Wed Jan 13 20:25:49 2010
URL: https://code.ros.org/trac/wg-ros-pkg/ticket/3558

ahendrix commented 11 years ago

[watts] We're having a problem in the left wrist on PRE. The left motor appears to glitch every once in a while. It may be jumping encoder counts and causing the assert to fail.

This is a problem, but shouldn't crash realtime/pr2_etherCAT.

ahendrix commented 11 years ago

[sglaser] I feel pretty ok about this assert, since it should only show up in the case of severe hardware failure, but I changed it to report an error anyways. It will break realtime, and it's extremely likely that calibration won't succeed anyways, but it won't crash.

r29145 which will be pr2_controllers-0.3.1

ahendrix commented 11 years ago

[watts] The problem with breaking realtime is that once we do it, none of our debugging or diagnostic tools work. Having a ROS_ERROR there is probably the best thing we can do.

Is having a calibration "fail" under consideration at all?