2026 RCJ Rescue Rule Changes

Dear RCJ Rescue Community,

The RCJ Rescue Committee is excited to share with you the upcoming 2026 rule changes that apply to all the different RCJ Rescue competitions, RCJ Rescue Line, RCJ Rescue Maze & RCJ Rescue Simulation. We are working on incorporating this to the official rules as soon as possible. In the meantime, we want to let you know to start preparing, discuss the changes with your teams and provide any early feedback to the committee over this forum. Once again, this year’s changes required a lot of discussion, understanding the latest technology advancements, analyzing the team’s performance during the 2025 competition and determining the goals of the RCJ Rescue competition in an international setting. Having a community discussion over the proposed changes will help us make sure the final version of the rules will be in the best shape and form to have a competition that encourages innovation and learning among their competitors in a fair scenario, like we learned from last year’s changes.

RULE CHANGES

A.1) AI-Based Solutions

Last year we initially proposed to allow AI-Based solutions without restriction including the incorporation of AI-based sensors like the Pixie Cam or Husky Lens. Talking with the community we learned that there was a big concern regarding this change and we decided to not implement it in 2025, instead we added a definition for tool, calibration and development in the efforts to encourage teams to do their own intelligent solutions to solve the different challenges. Unfortunately, as we later learned, the rule of not allowing all AI solutions was really hard to enforce. Each team had a different understanding of what development or innovation means. For example, some teams spent hundreds of hours training an existing AI model and did not consider this to be calibration. That’s just one example, we attempted to further bucketize the different AI solutions we saw were submitted for the competition but there isn’t a clear line of what is actual development, what is really innovation and what effort should value more than others for the robot development and challenge resolution.

That being said, the 2026 Committee decided to push once more to allow all AI based solutions without restriction, including AI-based sensors. The teams are still required to fully understand and explain the solutions they are implementing, including how they obtained the data utilized to train the AI models, how the training process is executed and how they are iterating over time to improve on the desired outcome.

To overcome some of the shared concerns from last year, we decided to make rule changes to each league to make sure that having an AI solution won’t be the core of the team’s results. If a team wants to succeed in the different challenges, existing AI solutions should not give a big edge to the team’s and if they want to succeed they will need to find ways to incorporate these AI solutions with their own developments and ensure a higher performance (and hence higher score). Before providing feedback on this topic, I encourage each of you to review each league rule changes to see how we are going to achieve this goal.

A.2) Superteam Changes

We are also making some changes to the super team challenge. We will be more transparent about the super team process in the rules so the teams know how to prepare for this part of the event. As a reminder, Superteams is a core not optional activity to perform in the RoboCupJunior competition and we want teams to be even more excited about it. We believe that the learning opportunities and connections that this piece of the competition brings are super valuable and we would like to keep expanding on those for all teams that will come with us to South Korea!

Another Superteam change is to explicitly ban bringing an additional robot to share with your Superteam partner. This has not been allowed so far, but we want to highlight in the rules the importance of sharing and supporting each team (not only your Superteam partner) while also making sure teams understand the best way to achieve this is to provide support and share knowledge and not do the work for the other team.

A.3) Multiple robot clarification

There has been some confusion for a while regarding the usage of multiple robots. While the different RCJ Rescue challenges are intended to be solved by a single robot and we consider them challenging enough for a team of maximum 4 people to solve it, we have seen teams that are interested in building multiple robots as a safety mechanism in case the robot stops working as expected in the middle of the competition. The existing rules do not clarify our position about this explicitly and sometimes teams can get confused reading between lines to understand what is allowed vs what is not. This rule clarification won’t change our position in the usage of multiple robots but will explicitly call it out. This means, explicitly mentioning that teams are allowed to bring two or more robots to the competition if and only if:

  • The robots are identical, using the exact same components and code.
  • The team is using only one robot at the time (including in the test fields).
  • If the team wants to switch robots, the team needs to get approval by the committee. This is to make sure there is a proper justification to switch (for example, robot failure, not enough battery, etc).
  • Teams cannot decide which robot to use based on the field their run is going to be.

Please let us know your feedback as soon as possible as we want to release the official final rules in the upcoming months.

For the specific league changes, please take a look at these posts:

Best,

Diego Garza Rodriguez on behalf of the 2026 RCJ Rescue Committee

5 Likes

Dear Rescue Committee,

Does the ruling described in this post

still apply? In other words, can we score the full amount of points using our current approach of utilizing a pretrained YOLO object-detection model (e.g., YOLOv8) trained on images we took ourselves?

Best regards,
Team BioBrause

1 Like

Hello @moritz_biobrause !

In our proposal for the 2026 rule changes we aren’t categorizing the different solutions as we did for 2025 based on rule 4.1 and the forum post you’re announcing. Instead, we decided to make rule changes to balance the competition and allow any use of AI tools and models.

That being said, you will continue to get full points for using pretrained YOLO object-detection model.

Thanks,

Diego Garza Rodriguez
2026 Committee

2 Likes

Dear Rescue Committee,

does the rule change officially permit the use of large language models (e.g., ChatGPT) during the development process, or are there explicit limitations or regulations limiting their use?

Best regards,
Team BioBrause

1 Like