Coding period week 7

1 minute read

Week 7

Last week I started upgrading the gym gazebo library that only has support for ubuntu 18.04 with melodic. As a result, to be compatible with behavior studio that works with ubuntu 20.04 and noetic the library needs some refactoring, since it is a key piece for the project to have it working in ubuntu 20.04.

After some struggle with the dependencies at gym gazebo, it works for the formula 1 model, although at the moment it is not fully integrated with behavior studio so it is working independently for now. Also more work has been done decoupling the formula 1 from old dependencies but it is not finished yet.

Gym Gazebo for ubuntu 20

At first I tried to update each dependency to noetic, but that would needed a lot of refactoring, some dependencies broke because in ubuntu 20 there was new versions for others tools so I could not make them compile.

Ubuntu 20 supports PCL 1.10 (Point cloud library) as default, but gym-gazebo has the package ar_track_alvar (ROS wrapper for Alvar) as a dependency which only seems to work with PCL 1.8.1. [3]

I ended reviewing which dependencies are necessary for the project and got rid off the ones that are not being used for the formula 1 model.

After I manage to compile the library, I started updating some old python2 syntax, and merged gym-gazebo’s requirements with the requirements at Behavior Studio.

I tested that everything was working with CPU and GPU, so the following steps was to update docker images with this update and document the installation properly.

Formula 1 plugins

I started from minimal working formula 1 plugins so I can add more functions incrementally and making tests a long the way. This work is going to be hosted at the CustomRobots repository.

This work should be finished next week along with the integration of the gym gazebo library into the behavior studio project.

Week Highlights

References