eprllib.Agents.ActionMappers

ActionMappers

This module defines the classes and functions used to perform actions on the environment.

ActionMapper functions are used to adapt the action given by the neural network, normally an integer for discrete spaces and a float for continuous spaces like Box. The actions must be adapted to values required for the actuators in EnergyPlus. Each agent has the capacity to control one actuator.

ActionMapper must be defined in the EnvironmentConfig definition to create the environment and is called in the Environment.Environment class and used in the Environment.EnvironmentRunner class to transform the dict of agent actions to actuator values.

The module includes the following classes:

• BaseActionMapper: The base class for creating ActionMapper functions.

• ActionMapperSpec: which is used to specify the configuration of ActionMapper functions in the agent definition.

These classes are used in the EnvironmentConfig class to specify the actions that can be performed on the environment.

Modules

ActionMapperSpec	Specification for the ActionMapper functions This module defines the ActionMapperSpec class, which is used to specify the configuration of ActionMapper functions for agents in reinforcement learning environments.
BaseActionMapper	Action Mapper Base Class