Final implementation of mountain_car basic scenario (Month 13 - Second half)
The goals of this two weeks were:
- get a productive solution for the mountain-car problem in RL-Studio 0.1.0.
- Allig whith the rest of the JDeRobot team so we cooperate to evolve RL-Studio.
Regarding the mountain car adaptation/migration, the following actions has been performed to make the simple environment built in the previous two weeks work:
-
Stable robot was imported an modified and the environment was refined to make the simulation controlable.
-
The statistics are propperly displayed while the simulation is run (in a different thread)
-
After lot of work and tunings to make it learn, the brain is finally learning!!
-
A new mode to execute actions manually was created (pressing 0 and enter the robot goes in one direction and pressiong 2 and enter it goes in the other direction). The name of the scrip is manual_runner.py
Note that, when the new version of RL-Studio evolved by Pedro is updated, this and the robot mesh implementations will be adapted and a new video to show this and the next two weeks work will be provided in “projects” section of this blog
Lectures
No new lectures were used this two (actually three) weeks
Lab work
- This week, the laboratory work consisted of keeping implementing the adapted/migrated to Rl-Studio mountain-car problem. Finally successfully!!!!.