I am reviewing some CoRL (Conf. on Robot learning) and Humanoids papers (6 papers total). None of them are reporting tests with a real robot (only simulation) => When did the robotics community stop caring about actual robots? these papers should be submitted to SIGGRAPH!
-
Show this thread
-
Replying to @jb_mouret
I disagree. There are now realistic large scale simulation tools that enable us to experiment in a completely different way compared to getting measurements with a real robot. algorithmic questions on how to design robust sample-efficient interactive learning is best studied here
4 replies 1 retweet 29 likes -
Replying to @AnimaAnandkumar
First, none of these papers is using "realistic large scale simulations". Most of them are using the "mujoco tasks" with only a few dof, no environment and no realistic actuator/sensor. Second, I love simulations as tools, but the results should still be validated in reality.
1 reply 0 retweets 9 likes -
Replying to @jb_mouret
Asking one research team to do is not fair. Focus on scientific contributions of the paper rather than a strict rule that all papers should be on a real robot. good algorithmic development it should be accepted. That's how we enable interdisciplinary research in robotics.
1 reply 0 retweets 17 likes -
Replying to @AnimaAnandkumar
Of course we focus on the scientific contributions. But a robotics paper is much more convincing if it actually works on a real system since it shows thay the result does not depend on the specific assumptions of a specific simulator.
1 reply 0 retweets 6 likes -
Replying to @jb_mouret @AnimaAnandkumar
I think the problem of transferring learning from simulation to reality shouldn't be neglected for a robotics conference. Even the most robust algorithm in simulation might fail in reality. Saying it's SEP is a bit of a cop out, because it might never work for real.
1 reply 0 retweets 8 likes
Agree that as a community we should not lose track. But asking every single paper to be only about that is short sighted. There are so many hard problems to tackle when it comes to robot learning. Open minded thinking will bring more people in and creativity will flourish
-
-
This Tweet is unavailable.
-
Thanks Amy. I certainly hope this is discussed. I agree that simulations don't complete the picture. But blindly rejecting such papers holds research back. Simulations are one step in the process and they should be treated as first class citizens
0 replies 0 retweets 4 likes - Show replies
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.