Qualifying Stage is now available!

[PuzeLiu]

We are glad to share that our Air Hockey Challenge has been accepted into the NeurIPS competition track. NeurIPS has long been recognized as a leading conference in the field of machine learning and AI from around the world. The addition of the Air Hockey Challenge to this esteemed lineup further underscores the growing interest in combining AI and robotics in real-world applications.

At this point, we are glad to release the qualifying stage. In this stage, you will work on a 7 DoF KUKA iiwa robot on three tasks: hit, defend, and prepare. You will train your agent to achieve the tasks of different difficulties while ensuring the deployability requirements.

Based on the feedback from the participants and the reviewer. We have made the following updates:

[Updates]

1. We provide several action interfaces to increase the flexibility of controlling the robot. Now you can use a different order of polynomials or send a trajectory buffer at each time step. Details here: https://air-hockey-challenges-docs.readthedocs.io/en/latest/qualifying.html#action-interface

2. We have extended the deadlines for each phase. We hope this helps to reduce the time pressure and allows for better results. The new schedule of the challenge can be found here: https://air-hockey-challenge.robot-learning.net/important-info

3. We have noticed the difficulty of satisfying the jerk requirement. Also, depending on the interpolation method, computing the jerk is not straightforward. We decided not to consider the jerk in the rest of the competition so that participants could focus on maximizing the agent's performance.

4. Teams at the Improvable or Deployable level are qualified to participate in the tournament stage. The maximum number of teams is 16, which will be determined based on the ranking of success rate in the qualifying stage.

[Important Info]

You can check out the branch to access the 7DoF environment:
https://github.com/AirHockeyChallenge/air_hockey_challenge

The documentation for the qualifying branch is also available:
https://air-hockey-challenges-docs.readthedocs.io/en/latest/qualifying.html

Model mismatches between the simulator and the evaluator will degrade the performance of the agent and even violate constraints. We strongly encourage participants to submit their agents multiple times to collect data sets from the evaluators and adjust performance. A leaderboard will showcase the best-performed records.

If you have any questions or concerns, please do not hesitate to contact us. We are here to support you in any way we can. We can't wait to see your amazing submissions!

Best regards,
The organizers

Website: https://air-hockey-challenge.robot-learning.net/
Documentation: https://air-hockey-challenges-docs.readthedocs.io/
GitHub: https://github.com/AirHockeyChallenge/air_hockey_challenge