search

osrf / srcsim

Space Robotics Challenge
Other
9 stars 3 forks source link

Latest srcsim leads to more surprising falls #254

Open osrf-migration opened 7 years ago

osrf-migration commented 7 years ago

Original report (archived issue) by Jeremy White (Bitbucket: knitfoo).

The original report had attachments: walkfail1.tgz

We have not had time to really study this, and most likely will not, but I wanted to record this so others can look for correlations.

We have all of task 1 automated to completion, and have for some time now. We run this process overnight on multiple machines in batches. Prior to yesterday, we had about a 75% success rate, with the bulk of the failures coming in the form of harness failures, gazebo crashes, crazy satellite dishes, or the occasional bug in our code. However, suffice it to say that we have an established, and 'routine' set of behaviors.

Last night we ran 13 such runs against the latest srcsim. 3 of those had a novel failure - Val simply collapsed during routine walks. That's a new and unusual failure pattern, and all we can imagine is that one of the many patches that went in last night introduced it.

Nicely, the runs were faster (25 minutes versus 35 minutes). And the harness and other failures did decline.

There are too many variables and too little time for us to be able to try to triage this any better. I'll see if I have logs I can upload.

osrf-migration commented 7 years ago

Original comment by Jeremy White (Bitbucket: knitfoo).

An example of a new walking failure; state.log and log files included.

osrf-migration commented 7 years ago

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

I'll take a look at this today. If you have a chance to rosbag record the force torque sensors, I think that is also helpful to see, since the IHMC controller uses those sensor readings quite a bit for walking.