Robot falls while walking in the habitat

[osrf-migration]

Original report (archived issue) by Víctor López (Bitbucket: Victor Lopez).

---

I opened a new issue because #250 refers to carrying tools and the overlapping collisions of the ground, and this one is not related to that.

The robot when walking in the habitat is much more prone to falling than when walking outside the habitat. The feet sleep and continuously, sometimes the robot recovers, sometimes it does not.

The video below is recorded using srcsim 0.8.0, with non-overlapping collisions on the habitat and not carrying tools. Using the keyboard teleop script provided by the organization and no extra software beside what's started using roslaunch srcsim unique.launch init:=True
https://youtu.be/PH85ipj2KzI

I recorded a rosbag of all ihmc topics and is here: https://drive.google.com/file/d/0By94Clyfum3XTFQxYmZSdHp3QjQ/view?usp=sharing

We have seen this in Round 1 and 2 of the finals, we have tried it 11 times and the robot has fallen in a similar fashion in all 11 times. We have managed to reach the table and pick the tool in ~5 of those tries, but haven't been able to reach the wall.

We haven't reached that with Round 3-5, but I believe the same will happen.

This is a blocking issue, I understand no code is going to be changed now that some teams have completed their runs, but would like to know how the organization will handle this.

We have picked up the detector in many runs, we would be able to detect the leak if we managed to get to the wall, we would get 15 points for that, or not, but there's no way to know it.