Module moog.tasks.stay_alive

Task that gives reward at regular intervals during task.

Expand source code 
"""Task that gives reward at regular intervals during task."""

from . import abstract_task


class StayAlive(abstract_task.AbstractTask):
    """StayAlive task. In this task a reward is given at regular intervals."""

    def __init__(self, reward_period, reward_value=1.):
        """Constructor.

        Args:
            reward_period: Int. Number of steps between each reward.
            reward_value: Scalar. Value of reward given.
        """
        self._reward_period = reward_period
        self._reward_value = reward_value

    def reset(self, state, meta_state):
        pass

    def reward(self, state, meta_state, step_count):
        """Compute reward."""
        del state
        del meta_state
        
        if (step_count + 1) % self._reward_period == 0:
            reward = self._reward_value
        else:
            reward = 0
            
        return reward, False

Classes

class StayAlive (reward_period, reward_value=1.0)

StayAlive task. In this task a reward is given at regular intervals.

Constructor.

Args

reward_period
Int. Number of steps between each reward.
reward_value
Scalar. Value of reward given.
Expand source code 
class StayAlive(abstract_task.AbstractTask):
    """StayAlive task. In this task a reward is given at regular intervals."""

    def __init__(self, reward_period, reward_value=1.):
        """Constructor.

        Args:
            reward_period: Int. Number of steps between each reward.
            reward_value: Scalar. Value of reward given.
        """
        self._reward_period = reward_period
        self._reward_value = reward_value

    def reset(self, state, meta_state):
        pass

    def reward(self, state, meta_state, step_count):
        """Compute reward."""
        del state
        del meta_state
        
        if (step_count + 1) % self._reward_period == 0:
            reward = self._reward_value
        else:
            reward = 0
            
        return reward, False

Ancestors

Methods

def reward(self, state, meta_state, step_count)

Compute reward.

Expand source code 
def reward(self, state, meta_state, step_count):
    """Compute reward."""
    del state
    del meta_state
    
    if (step_count + 1) % self._reward_period == 0:
        reward = self._reward_value
    else:
        reward = 0
        
    return reward, False

Inherited members