An Explorer Env (Incomplete)¶
In this tutorial we’re going to take the minimal environment we studied earlier and build it into
- Compared to the minimal env, the explorer env
Has compound RGB-depth-IMU observations.
Has movement with momentum.
Outputs rewards for exploring.
Resets agents when they’ve lived too long.
You can see the final explorer code here.
Like in the second half of the minimal environment tutorial, when discussing the construction of a large class it isn’t feasible to copy-paste the entire definition every time. Instead, the snippets here are going to be small ‘experiments’ that demonstrate how a particular part of the class works.
The first thing to do is to make sure you can actually run the complete explorer environment. Open up a notebook or IPython console and
TODO-DOCS Rest of the explorer tutorial