Location
Badges
Activity
Ratings Progression
Challenge Categories
Challenges Entered
Identify user photos in the marketplace
Latest submissions
Robots that learn to interact with the environment autonomously
Latest submissions
See Allgraded | 87061 | ||
graded | 87047 |
Sample-efficient reinforcement learning in Minecraft
Latest submissions
Participant | Rating |
---|---|
nguyen_thanh_tin | 0 |
Participant | Rating |
---|
REAL 2020 - Robot open-Ended Autonomous Learning
Wrappers using / observation space access
About 4 years agoAbout wrappers: It was just a suggestion, no problems
About observation space: Thank you)
About βobject_positionβ: I mean βobject_positionβ space.shape vs βobject_positionβ observation.shape.
Environment observation space is taken from βrobotβ attribute - Kuka class. Kukas observation space is Dict space. There is key βobject_positionβ and it corresponds to Dict space with keys [βtomatoβ, β¦]. This spaces (βtomatoβ-space and etc.) are Box spaces with shape (7,) (real_robots/envs/robot.py, line 75). But environments [βstepβ, βresetβ] methods returns observation where observation[βobject_positionβ][βtomatoβ].shape is (3,), because get_position() is called instead of get_pose() (real_robots/envs/env.py, line 234).
Wrappers using / observation space access
Over 4 years agoAlso environments βobject_positionsβ spaces shape differs from corresponding shape in observation: (7,) vs (3,). I guess problem is in get_position() method calling (returns only coordinates) instead of get_pose() (returns coordinates and orientation).
Wrappers using / observation space access
Over 4 years agoHello!
Is there any way to use wrappers? There are None values (for βgoal_maskβ and βgoal_positionsβ keys) in observation dict in R1-environment. It can be solved with adding zero values for this keys to 93 line in real_robots/env.py:
self.goal = Goal(retina=self.observation_space.spaces[
self.robot.ObsSpaces.GOAL].sample()*0)
or with use of wrappers.
Also it can be useful if observation_space also was provided to controller (for nn model defining and etc.). In my code I got information about observation_space from Kuka class, but it is not the most elegant way)
Baseline question
Over 4 years agoHello! Question about βpercentage_of_actions_ignored_at_the_extremesβ parameter.
As I understand this parameter allows us to drop the least relevant distances. Should there be np.linspace(actions_to_remove, len(self.actions) - 1, β¦) or np.linspace(0, len(self.actions) - 1 - actions_to_remove, β¦) instead of np.linspace(actions_to_remove, len(self.actions) - 1 - actions_to_remove, β¦) in abstractor.py:
for i in range(condition_dimension):
sup = ordered_differences_queues[i].get_queue_values()
for j in np.linspace(actions_to_remove, len(self.actions) - 1 - actions_to_remove, config.abst['total_abstraction']).round(0):
self.lists_significative_differences[i] += [sup[int(j)]]
?
NeurIPS 2019 : MineRL Competition
New obtaindiamond
About 5 years agoThere are normal rewards in the latest updates (once per item except logs). But you havenβt changed docker and submissions are evaluated with βreward bugsβ.
[Announcement] Submissions for Round 1 now open!
Over 5 years agoQuestion about deadline of first round: https://www.aicrowd.com/challenges/neurips-2019-minerl-competition there is said that 1 round finishes in 48 days, but it differs from date in βimportant datesβ (22 september). When first round finishes?
How is the "reward" on leaderboard page computed?
Over 5 years agoAlso it looks like it is βDenseβ environment, because using evaluate_locally.sh script weβve got reward for every crafted item, and after replacing βObtainDiamondβ with βObtainDiamondDenseβ weβve got reward only once per item.
Cartesian space question
About 4 years agoHello. Weβve found cartesian space slows down fps. For example on my PC using βmacro_actionβ and βjointsβ action spaces environment could make around 1000 steps per second. But βcartesianβ slows down to 100 steps per second.
The reason is inverse kinematics calculation. Every environment step is simulation step, so to change arm pose in βjointsβ or βcartesianβ spaces you should send the same action for 100-500 steps and the same inverse kinematics calculations are performed 100-500 times. To speed up actions in βcartesianβ space action caching can be used (as in βmacroβ space). Also βgripper_commandβ is ignored in βcartesian_spaceβ.