2026 RCJ Rescue Maze Rule Changes

[Dieguinilombrin]

Dear Teams,

After carefully reviewing and discussing the experiences and feedback from last year, we have introduced the proposal of changes to the Rescue Maze 2026 Draft Rules. Although the discussions took some time, we truly hope these updates will bring new and exciting challenges for all of you.

In the 2026 Maze draft, we have outlined six key topics that we plan to revise in the final version of the rules. Before making any final decisions, we would like to hear your opinions, comments, and questions regarding these proposed changes.

Please take a moment to read through the document and share your feedback.
We wish every team a successful preparation for the upcoming season. We look forward to meeting you all in Incheon, South Korea!

Topic 1: Victim Identification and AI Strategy

Here are the proposed changes:

• Coloured Victims: Color victims (green, yellow and red squares) will be removed.
• Visual (Letter) Victims: Visual victims will replace the colour victims in the score. The letters will change to the greek alphabet (Φ, Ψ, Ω).
• NEW: “Cognitive Target”: This new target is designed to test identification, calculation, and decision-making.

1. Identification:

The robot must identify a circular target with a 5 cm diameter. This target is composed of 5 concentric rings (each 0.5 cm wide), and each ring will have one of 5 possible colors: Black, Red, Yellow, Green, or Blue.

2. Calculation:

The robot’s primary task is to “read” the target and calculate a total value.

First, each color corresponds to a numerical value:

• Black = -2
• Red = -1
• Yellow = 0
• Green = 1
• Blue = 2

Adjacent rings of the same color are not merged. The robot must always consider each of the 5 rings separately and sum the value for all 5 rings, regardless of whether colors repeat.

Example 1:

1. Rings (from center → outwards): Red | Yellow | Yellow | Blue | Blue[image]
2. The robot reads five separate color values.
3. Calculation: Value(Red) + Value(Yellow) + Value(Yellow) + Value(Blue) + Value(Blue)
4. Final sum: (-1) + (0) + (0) + (2) + (2) = 3
5. The robot must act based on this total (Sum = 3 → false target).

Example 2:

1. Rings (from center outwards): Blue | Blue | Blue | Black | Black[image]
2. The robot reads all five rings individually.
3. Calculation: Value(Blue) + Value(Blue) + Value(Blue) + Value(Black) + Value(Black)
4. Final sum: (2) + (2) + (2) + (-2) + (-2) = 2
5. The robot must act based on this total (Sum = 2 → blink the LED and drop two rescue kits).

3. Action:
The robot must perform a specific action based on the final calculated sum:

• If Sum = 0: The robot identifies the target (by stopping for 5 seconds and blinking the victim identification LED)
• If Sum = 1: The robot identifies the target (by stopping for 5 seconds and blinking the victim identification LED) and drops up to 1 rescue kit
• If Sum = 2: The robot identifies the target (by stopping for 5 seconds and blinking the victim identification LED) and drops 2 rescue kits
• If Sum is < 0 or > 2: The target is considered “false.” If the robot identifies it (by stopping for 5 seconds and blinking the victim identification led), it counts as misidentification resulting in -5 points

4. Scoring (Cognitive Target):

• Identifying a cognitive target on a linear tile awards 10 points, while identification on a floating tile awards 30 points.
• Correctly placing one rescue kit awards 10 points, and correctly placing two rescue kits awards 30 points.

Example Scoring:

• Correctly flashing for Sum=0 (on floating wall): 30 points.
• Correctly deploying 1 kit for Sum=1 (on floating wall): 30 (base) + 10 (1 rescue kit) = 40 points.
• Correctly deploying 2 kits for Sum=2 (on floating wall): 30 (base) + 30 (2 rescue kits, 10 + 20 points) = 60 points.

Cognitive Target Illustration

The Cognitive Target is a circular pattern with a 5 cm diameter, consisting of up to five concentric rings.
The innermost circle has a diameter of 0.5 cm, and the width of each subsequent ring increases by 0.5 cm, resulting in ring diameters of 0.5 cm, 1.0 cm, 1.5 cm, 2.0 cm, and 2.5 cm.
Adjacent rings of the same color are not merged. The robot must always consider each of the 5 rings separately and sum the value for all 5 rings, regardless of whether colors repeat.

Topic 2: Blue Tile Scoring

We propose implementing a new scoring for blue tiles. A team will be awarded 30 points the first time their robot successfully visits a unique blue tile.

However, the score for a specific tile is reduced upon revisits. For every subsequent time the robot visits that exact same tile, the total score awarded for that specific tile will be reduced by 10 points.

The minimum bonus for any single tile is 0 points; it cannot become negative.

Example (Tracking a single, specific blue tile):

• 1st entry onto the tile: The team is awarded 30 points for this tile.
• 2nd entry onto the same tile: The total score for this tile is recalculated to 20 points (30 - 10).
• 3rd entry onto the same tile: The total bonus for this tile is recalculated to 10 points (20 - 10).
• 4th entry onto the same tile: The total bonus for this tile is recalculated to 0 points (10 - 10).
• 5th entry onto the same tile: The total bonus for this tile is recalculated to 0 points (0 - 10 is -10, but it can not go negative).

Any further entries onto this specific tile will also yield 0 points. If the robot later enters a different, previously unvisited blue tile, it would receive a new, separate 30 points for that second tile.

Topic 3: Exit/Reliability Bonus Rework

We propose to change the calculation of the Exit and Reliability Bonus. Instead of being based solely on points from identified or rescued victims, we want the bonuses to reflect the robot’s overall performance.

This change aims to make the Exit and Reliability Bonus more balanced rewards for a complete and successful run, rather than linking them exclusively to victim-related tasks.

The calculation of the bonuses could look like this:

(RELIABILITY BONUS) = (SVI) × 10 + (SRD) × 10 + (MBT) × 10 - (LoP) × 15
MBT … Maximum BlueTiles navigated: Gain 10 points for every blue tile which was only visited once.

(SUCCESSFUL EXIT BONUS) = (SVI) × 10 + (MBT) × 10 + (SSN) × 5 + (SRN) × 5
SSN … Successful Stair Navigation
SRN … Successful Ramp Navigation

Topic 4: Rescue Kit Reduction

We propose to reduce the total number of available rescue kits from 12 to 8. We believe this change will place a greater emphasis on resource management during the run.

We are also revising the points awarded for placing rescue kits next to victims.

• Correctly placing one rescue kit earns 10 points.

• Correctly placing two rescue kits earns 30 points.

This means:

• If a victim or target requires one rescue kit, the robot can earn 10 points by placing it correctly.

• If a victim or target requires two rescue kits, the robot can earn up to 30 points for placing both correctly.

• If the robot places only one rescue kit at a location that requires two, it will receive 10 points.

Topic 5: Field Marking Prohibition

We propose to clarify and strictly enforce the rule: The robot (and team) is not allowed to mark the field.

We have had experiences where teams creatively marked the start tile as a way to help their robot secure the Exit Bonus. We believe this practice must be explicitly clarified and disallowed for two main reasons:

1. Providing External Information: A team marking any part of the field constitutes intentionally providing the robot with information about the environment, which is not permitted.
2. Spirit of the League: This practice goes against the core concept of Rescue Maze, which is to encourage teams to develop robots that can autonomously map and navigate an unknown space. Using pre-placed markers circumvents this fundamental challenge.

Topic 6: Field Clarifications/Changes

• The definition of walls also include the support structure on the ends (e.g. item profiles, pillars, …)
• To make the construction of fields easier and due to the walls not being indefinitely thin, the pathway between two opposite walls is 28cm wide (±10% tolerance).

Best,

Diego Garza Rodriguez on behalf of the 2026 RCJ Rescue Committee

[PeterParker]

1- Thank you for sharing the six new discussion points.

2- Our first official competitions will be held in March. The competition judges will apply the rules valid at the end of January 2026. We do not yet know what they will decide for the national competitions to be held in April. The risk is that we will then participate in the European and World Championships (if applicable) with different final rules. For example, regional and national competitions with the 2025 final rules and European and World Championships with the 2026 final rules. In other words, this delay in presenting the 2026 final rules creates considerable difficulties. Teams may be forced to change their hardware and software, even significantly, in a few months. Furthermore, as of today, we still do not know exactly how to design the robot. I leave the conclusions to you.

3- Undoubtedly, the 30-degree incline staircase and the new cognitive target, together with the new rules on blue tiles, represent an amazing challenge. However, I fear that continuing to increase the difficulty of the maze could lead to very few teams being able to compete decently and others abandoning the competition altogether because it is too difficult.

4- It would be easier to change the rules over the years in order to stimulate the teams with small changes, but without this process leading to overly complex solutions.

I look forward to hearing the opinions of my colleagues and the decisions of the committee.
Thank you.

[IGSKaufungen]

To topic 6: Field Clarifications/Changes:

Previously, maze arenas were often built this way:

Because the rules said: All tiles are defined as a 30 cm x 30 cm space. And if the walls have a thickness of 2cm, the Space between the walls are 30cm!
But in my opinion, that means that all fields must be 30cm x 30cm, regardless of whether a wall is standing there or not.
Therefore, I conclude that the Maze Arena must be sized as such:

This, of course, also includes the new rule.
Only: We won’t build any new arenas now, at least not in Germany.
But with a tolerance of 10%, the space between the walls is still legal.

Thank You.
Have a nice day
Andreas

[PeterParker]

Topic 5: I am completely in favour.

In some international championships, I noticed that some teams placed a yellow piece of paper under the robot, pretending it was part of the robot itself. In this way, they left the piece of paper on the starting tile, making it very easy to recognise and win the exit bonus. There was no need to map the route. In my opinion, the old rules already prevent this move, but the paragraph is not very clear and the judges interpreted it in such a way that this practice, which in my opinion undermines the spirit of the game, was actually allowed.

It is not clear to me whether this new rule that you want to introduce in topic 5 prevents practices such as the one I have just mentioned.

[jomue]

First of all, thank you for sharing the proposed rule changes. Overall, I really appreciate them—especially the introduction of the new victims. However, I have concerns about two parts: the reduction of Rescue Kits and the revision of the scoring system.

Regarding the Rescue Kits, I believe that storing and managing them has always been an interesting hardware challenge. Reducing the number of kits simplifies this aspect significantly, which feels like a loss rather than an improvement.

More importantly, I disagree with the new point system. I understand the intention of placing greater emphasis on resource management during the run, but I think this rule introduces a substantial amount of randomness into the competition. To handle the new scoring system, teams would likely consider two strategies (or variations of them):

1. Always place a Rescue Kit as soon as the robot detects a victim.
   This avoids running out of time later but risks losing points if you run out of kits needed for victims requiring two kits.
2. Only place kits for the two-kit victims initially, then deliver the remaining kits to one-kit victims after exploring the entire maze.
   This may yield more points, but it carries the risk of running out of time or being forced to cross blue tiles again to reach earlier victims.

Because the size and complexity of the maze are unknown, the second strategy is inherently risky. Teams might try to mitigate this by deciding before each run which strategy - or variation of a strategy - to use. However, this could easily be interpreted as pre-mapping, and in my opinion goes against the spirit of the competition. Furthermore there is no realistic way to enforce or detect whether a team adjusted their strategy based on the specific arena or simply because they believed it was generally better.
Lastly, this change also adds a significant amount of complexity. Combined with the new Blue Tiles and the new Victims/Targets, the overall difficulty would increase dramatically. As others have already pointed out, raising the complexity this much in a single year may not be beneficial for the league.

I’m interested to hear what the rest of you think about these concerns.

Best Regards
Jonas

[zsofi_kp]

Dear Committee,

We would like to thank you for sharing the planned rule modifications with the teams. We find the proposal creative and innovative. The new ideas are also useful because they motivate teams that have participated in the category for a longer time as well. We are pleased to see your efforts to create a balance between the use of AI and the development of custom-made software. This is an exciting challenge that we look forward to.

Our observation is that the diameter of the new cognitive victims is 5 cm. This is 1 cm larger than the size of the previous victims. If the new 5 cm diameter victims are placed according to the previous rules, their top edge could be at 7 cm + 5 cm = 12 cm, and their bottom edge at 7 cm – 5 cm = 2 cm in height. Considering the size of the robots and the width of the corridors, we believe this may make accurate detection of the victims more difficult.
Based on the draft, we do not yet know whether the committee is planning any rule adjustments to ensure that the victims will continue to be placed within the 3 cm to 11 cm height range.

Respectfully,
Team Lightning,
Kiss Zsófia

[elizabeth.mabrey]

topics 1: Cognitive target

I think the scoring combination will need clarification:

1. “blink” - how many? intervals?
2. “led” - any single color of led? any onboard led… or onboard pixy as long as it is visiable?
3. scoring -
   • will each recognition be given partial scores or taking “all or nothing” approach.
   • if team may get partial score (e.g. 5 secs stop + blink, but did not drop correct # of kits)? It partial scores is given, I recommend to write out a table to clarify the scoring variation. I was doing it… but need to know how this really works first.
   • especially the clause - "…Correctly deploying 2 kits for Sum=2 (on floating wall): 30 (base) + 30 (2 rescue kits, 10 + 20 points) "… what is the extra 10 coming from?
   • or something as simple as did not stop enough time, etc.

Also, i think there is a typo “… The innermost circle has a diameter of 0.5 cm , and the width… etc.”. Perhaps it is trying to say:

"The outermost circle has a diameter of 5 cm, consisting of up to five concentric rings.
The innermost circle has a diameter of 1 cm, and the diameter of each subsequent outer ring increases by 1 cm, resulting in rings with diameters of 1 cm, 2 cm, 3 cm, 4 cm, and 5 cm.

---

I echo PeterParker’s point 3. Especially when there may be speed bump on the ramp.

[elizabeth.mabrey]

About topic 5 - leaving marker.

As long as the robot autonomously dispenses paper markers while navigating, this action should not be considered pre-mapping, right?

my humble opinion: In a real-world scenario, when a robot—or even a person—navigates through an unfamiliar environment such as a forest, it must find creative ways to retrace its path efficiently. This represents logical problem-solving, not deception. Likewise, leaving markers autonomously during operation should be viewed as an example of intelligent design and adaptive reasoning, rather than prior mapping.

–Elizabeth Mabrey

[elizabeth.mabrey]

Topics 6:

I think there is a conflcting information which needs further clarification - “the pathway between two opposite walls is 28cm (± 10% tolerance).”

The rule said tiles about 30cm , but with ±10% error margin, ie. between 27 to 33cm. So, now, if it is a hallway underneath a bridge, it ends up to 25.2 to 30.8 cm according to what is said in the post - 28 ± 10%. Thus, effectively, the bot must not exceed 25cm. I am not trying to debate about this. Just thought this should be made clear in the rule.

[PeterParker]

Dear Elisabeth,
thank you for sharing your opinions and practical suggestions.
Regarding point 5, I will describe what I have sometimes seen in competitions, especially at the European championships. Some teams place their robot on the competition field with a sheet measuring approximately 20x20 centimetres, usually a bright yellow colour, under the wheels, saying that it is part of the robot itself. As soon as the robot starts moving, the sheet remains on the starting tile. This trick makes it easy to recognise the aforementioned box, making it much easier to obtain the exit bonus, especially during maze entry. The reason is obvious. This solution is less effective for the “standard” maze, of course. In my opinion, the use of this stratagem undermines the spirit of the challenge, which is to create an effective mapping system based on the management of data relating to the distances travelled by the robot and the position of walls and chasms. In my opinion, this practice results in a sort of pre-mapping, as immediately before the start, the robot is given information about the type of maze it will face. The competition judges, on the other hand, say that there is no pre-mapping, as the rules state that the competition starts as soon as the robot is placed inside the maze.

The race judges told me that they had not found a way to prevent this type of “trick” under the rules. In the end, various teams, realising the possibility of using the coloured sheet, tried to implement a solution, but generally too late compared to those who had adopted it immediately. I conclude by saying that this possibility should be eliminated. I look forward to hearing your opinions and suggestions. Thank you.

[Dieguinilombrin]

Hello everyone!

Thank you so much for the discussion, the committee is reviewing every comment and discussing possible fixes and concerns while keeping the original objectives. Will try my best to reply to all the comments in this post.

Final Rule Publication
We are aware of the challenges and concerns with the timing of publishing the final 2026 rules. The approach that we prefer to take in the RoboCupJunior Rescue Committee is to have lengthy discussion analyzing every proposed rule before sharing it with the community. While this takes some time, it helps us narrow down from 70+ rule proposals to the ones you’re currently reading about. Regarding the rules that will be used by local, regional and national competitions, please reach out to the tournament organizer as each have different timings and decide which rules to use (for example, some regions already hosted their national competition using the 2025 rules).

Rule Difficulty and Complexity
We hear you about the complexity of the Rescue Maze challenge. As some of you pointed out, our goal with the rule changes is to achieve two main things, give the competition organizers more flexibility to adapt to the competitor level in their regions to have enough tools to identify the best teams and allow give teams a resource management challenge. We have seen over the years that teams keep improving and taking the challenge to the next level, with competitors always aiming to build a robot that scores perfect runs. While we would love that, having the best teams consistently scoring perfect runs makes it really hard to determine who are the best teams. This year we want for teams to see teams prioritize what is most important for them while rewarding the most resilient robots. Taking a decision if its better to enter the dangerous zone, drop all your rescue kits on the first victims you identify or if its worth to passing over the same blue tile multiple times to explore the entire field are decisions that the teams need to make not only while designing their robots but adapting based on the robot capabilities getting closer to the competition. If a team is not sure their mapping accurate, they might about the blue tiles even before accessing them. If a team is not sure their robot is fast enough to explore the entire field and come back to drop rescue kits to the victims that give slighly less points, they can continue to priority dropping kits as soon as a victim is identified. We aren’t expecting to see many rounds, but we are sure these decisions will help shape the competition in a very rewarding way.

Pre-mapping Concerns
Similar to other pre-mapping concerns, we tackle this by reviewing team’s submission documents and holding in person interviews to properly understand robot behavior and reasoning behind those decisions. If a team changes prioritization based on the field they will be playing, that is considered pre-mapping. If the robot is able to adapt those decisions by itself it’s fine. If the team needs to adapt strategy mid competition, similar to other situations, they need to go to the committee and let them know of their change to make sure there is a valid reasoning behind it and be approved for such.

Victim Placement with respect to ground
Thank you for pointing that out! We did multiple tests with a physical robot before publishing the cognitive target victims. This helped us determine that a victim size of 5cm was needed for it to be more effective. We aren’t planning any changes at the moment to change the 7cm off the ground rule, but we will be running more tests to see if there are any concerns with respect to the edges. We will update in this forum if we decide to change this to adjust it so it’s always un the 3cm - 11cm range instead of 2cm - 12cm.

Cognitive Target Victims
These are just new victims, all other rules regarding identification and dropping kits haven’t changed. As explained in the initial post, if a team correctly identifies a victim, it will get the identification points. If it drops one kit when it shouldn’t, there is no penalty. If it drops one kit when it could drop two kits, it will only receive the points for dropping one kit.

We can change the wording in the final rules to make it more clear regarding the dimensions.

Tile / Path Dimension Changes
I want to start clarifying that the tile size won’t change, it is still defined as a 30 x 30cm tile. The change we are adding is to simplify building and constructing the maze field by noting the pathway between two opposite walls can be 28cm wide. Andreas image showcases this problem (in Andreas image, the poles used to add walls are 1cm wide). Looking at the first image, the upper left tile has a path of 31cm, while the upper middle tile has a path of 32cm (because there aren’t any walls which thickness reduces the pathway size). In most countries, the material used for the base can be found in size 120cm x 90cm, which further reduces the pathway size. We have seen regions where they have been using pathways of 28cm even though it wasn’t explicitly callout in the rules. This is help clarify this and allow more flexibility when constructing the areas, no need to redo any existing fields.

PS: We are also working on releasing a proposed way of constructing the Rescue Line and Maze fields to minimize the entry barrier to new regions, and these changes will allow to even reuse some of the material between Line and Maze for those purposes.

Marking the field by intentionally dropping components
The scenario that PeterParker mentions is the specific scenario we wanted to tackle, the dropping of a big paper sheet at the beginning of the tile by adding duct tape to the bottom of their robot and making it “fall” at run start. The committee likes the idea of incorporating a challenging way of carrying “mobile checkpoints” that the robot can intentionally drop and use as guidance, but we determined that the rule changes that would be needed to incorporate something like this while disallowing the tapped paper are considerable that will complicate the rules further. We are tracking this as a future addition to the rules, but not in the 2026 ruleset.

Please keep the discussion going, we are always happy to hear your opinions and comments on the topic.

Regards,

Diego Garza Rodríguez
2026 Committee