Agent

We provide a base class AgentBase with some utils functions to extract the desired state from the observation. You can inherit the base class and implement your onw method in the air_hockey_agent/agent_builder.py file. A Dummy Agent example can be found in Dummy Agent.

Load and Save Agent

We also provide a simple and effective way of save and load your agent. We extend the Dummy Agent example and set different type of variables. You can add these variables into saving list by calling self.__add_save_attr function.

The available methods are:

• primitive, to store any primitive type. This includes lists and dictionaries of primitive values.

• numpy, to store NumPy arrays.

• torch, to store any torch object.

• pickle, to store any Python object that cannot be stored with the above methods.

• json, can be used if you need a textual output version, that is easy to read.

• none, add the attributes, you can assign the values to the attribute later.

import numpy as np
import torch

from air_hockey_challenge.framework import AgentBase, AirHockeyChallengeWrapper


def build_agent(env_info, **kwargs):
    """
    Function where an Agent that controls the environments should be returned.
    The Agent should inherit from the mushroom_rl Agent base env.

    Args:
        env_info (dict): The environment information
        kwargs (any): Additionally setting from agent_config.yml
    Returns:
         (AgentBase) An instance of the Agent
    """

    return DummyAgent(env_info, **kwargs)


class DummyAgent(AgentBase):
    def __init__(self, env_info, value, **kwargs):
        super().__init__(env_info, **kwargs)
        self.new_start = True
        self.hold_position = None

        self.primitive_variable = value  # Primitive python variable
        self.numpy_vector = np.array([1, 2, 3]) * value  # Numpy array
        self.list_variable = [1, 'list', [2, 3]]  # Numpy array

        # Dictionary
        self.dictionary = dict(some='random', keywords=2, fill='the dictionary')

        # Building a torch object
        data_array = np.ones(3) * value
        data_tensor = torch.from_numpy(data_array)
        self.torch_object = torch.nn.Parameter(data_tensor)

        # A non serializable object
        self.object_instance = object()

        # A variable that is not important e.g. a buffer
        self.not_important = np.zeros(10000)

        # Here we specify how to save each component
        self._add_save_attr(
            primitive_variable='primitive',
            numpy_vector='numpy',
            list_variable='primitive',
            dictionary='pickle',
            torch_object='torch',
            object_instance='none',
            # The '!' is to specify that we save the variable only if full_save is True
            not_important='numpy!',
        )

    def reset(self):
        self.new_start = True
        self.hold_position = None

    def draw_action(self, observation):
        if self.new_start:
            self.new_start = False
            self.hold_position = self.get_joint_pos(observation)

        velocity = np.zeros_like(self.hold_position)
        action = np.vstack([self.hold_position, velocity])
        return action


if __name__ == '__main__':
    env = AirHockeyChallengeWrapper("3dof-hit")

    # Construct Agent
    args = {'value': 1.1}
    agent_save = build_agent(env.env_info, **args)

    print("######################################################")
    print("Save Agent Variables")
    print("######################################################")
    print("agent_save.primitive_variable: ", agent_save.primitive_variable)
    print("agent_save.numpy_vector: ", agent_save.numpy_vector)
    print("agent_save.list_variable: ", agent_save.list_variable)
    print("agent_save.dictionary: ", agent_save.dictionary)
    print("agent_save.torch_object: ", agent_save.torch_object)

    # The not_important variable will not be saved unless the full_save is set True
    agent_save.save("agent.msh", full_save=False)

    agent_load = DummyAgent.load_agent("agent.msh", env.env_info)
    print("######################################################")
    print("Load the Agent")
    print("######################################################")
    print("agent_load.primitive_variable: ", agent_load.primitive_variable)
    print("agent_load.numpy_vector: ", agent_load.numpy_vector)
    print("agent_load.list_variable: ", agent_load.list_variable)
    print("agent_load.dictionary: ", agent_load.dictionary)
    print("agent_load.torch_object: ", agent_load.torch_object)
    print("agent_load.object_instance: ", agent_load.object_instance)

    print("------------------------------------------------------")
    print("These variable will not be saved while full_save is False")
    print("agent_load.not_important: ", agent_load.not_important)

    print("------------------------------------------------------")
    print("These variable will be parsed from env_info:")
    print("agent_load.env_info.keys()s: ", agent_load.env_info.keys())
    print("agent_load.agent_id: ", agent_load.agent_id)
    print("agent_load.robot_model: ", agent_load.robot_model)
    print("agent_load.robot_data: ", agent_load.robot_data)

AgentBase

air_hockey_challenge.framework.agent_base