Making runfile configs accessable in custom Brain and Muscle
Improvement Issue
Description
The custom Brain
and Muscle
implementations may rely on configs denoted in the runfile or even experiment file, but currently they only dynamically get parameters give to them.
Which components will be improved?
All components from read-in of the runfile and possible modifications on it, down to the message passing/object creation to the custom implementations
Reason for improvement
- In the ARL architecture there is the need to access the Env (reward) and Environment objective functions to be able to call them.
- Additionally, for the transformation (with the NN2EQCDT algo) of the model the sensors have to be read out to identify its class and assign the corresponding constraints.
Order to Approach
- Pass through the configs on its own through palaestrai OR
- Write the configs in the db and make them transparently accessible by the custom implementations (as also discussed for the arsenAI loop) OR
- Backwards calling of functions (on higher level objects) to the the configs