class
MoveObjectsRewardA reward based on L2 distances to object/goal.
Methods
- def _get_count_coll_reward(self)
- Count-based collision reward
- def _want_count_coll(self)
- Check if we want to consider penality from count-based collisions
- def get_distance(self, task, distance)
- def get_metric(self)
- def reset_metric(self, *args, episode, task, observations, **kwargs)
- def update_metric(self, *args, episode, task, observations, **kwargs)
- def update_target_object(self)
- The agent just finished one rearrangement stage so it’s time to update the target object for the next stage.
Special methods
- def __init__(self, *args, **kwargs)
Data
Method documentation
def habitat. tasks. rearrange. multi_task. pddl_sensors. MoveObjectsReward. get_metric(self)
Returns | the current metric for Measure . |
---|