Separates the environment loop (data collection) from the learning algorithm.
Adds an entry-point for the project that runs the simulation
Adds typed configurations to reduce the likelihood of subtle type errors
These changes are essential to implementing more complex learning algorithms (and in the future planning algorithms as well). We can now implement learning algorithms that can be used on different parts of the agent that attempt to optimize for different objectives. It is also much easier to add to the configuration.
Accomplishes the following:
These changes are essential to implementing more complex learning algorithms (and in the future planning algorithms as well). We can now implement learning algorithms that can be used on different parts of the agent that attempt to optimize for different objectives. It is also much easier to add to the configuration.
Closes #12