renamed everything to step() if used in environment.step(). renamed everything...
16.06.
- introduced actions_are_relative in config
- made lineSpeed an output -> adapted everything necessary
- introduce control_manager and disturbance_manager instead of state_manager
- rename model_wrapper to output_manager
- name many things to step() -> beforehand get_controls() did not only return controls, get_reward() did not only return reward
- always use OmegaConf.to_container() instead of dict()
- rename to control_contraints_met and output_constraints_met -> not everything is safety -> safety could also be outputs, e.g. pressure or temperature -> same goes for requirements: control bounds may be more restrictive than what is possible for the machine
- restructured config
- outputs_are_latent and observation_noise_only in env_setup
- control_bounds and output_bounds first hierarchy level -> organized the same way with optional upper and lower bounds
- make it possible to determine outputs dependent on other outputs in output_manager.step()
- renamed to environment and gym_wrapper
- make reward_function a config
- stable baselines 3 works (also env_checker)
- actions_dict has same keys as controls
- change scenario manager process: one step()-method -> works like other classes -> maybe this way we can actually make a abstract class
- restructured the repo
- distributable
- include changes from review-ruben