#RoboCup2024 – day by day digest: 21 July

A break in play throughout a Small Measurement League match.

As we speak, 21 July, noticed the competitions draw to an in depth in an exciting finale. Within the third and last of our round-up articles, we offer a flavour of the motion from this final day. When you missed them, you’ll find our first two digests right here: 19 July | 20 July.

My first port of name this morning was the Normal Platform League, the place Dr Timothy Wiley and Tom Ellis from Group RedbackBots, RMIT College, Melbourne, Australia, demonstrated an thrilling development that’s distinctive to their workforce. They’ve developed an augmented actuality (AR) system with the intention of enhancing the understanding and explainability of the on-field motion.

The RedbackBots travelling workforce for 2024 (L-to-R: Murray Owens, Sam Griffiths, Tom Ellis, Dr Timothy Wiley, Mark Subject, Jasper Avice Demay). Picture credit score: Dr Timothy Wiley.

Timothy, the educational chief of the workforce defined: “What our college students proposed on the finish of final yr’s competitors, to make a contribution to the league, was to develop an augmented actuality (AR) visualization of what the league calls the workforce communication monitor. This can be a piece of software program that will get displayed on the TV screens to the viewers and the referee, and it exhibits you the place the robots suppose they’re, details about the sport, and the place the ball is. We got down to make an AR system of this as a result of we predict it’s so significantly better to view it overlaid on the sector. What the AR lets us do is mission all of this info reside on the sector because the robots are transferring.”

The workforce has been demonstrating the system to the league on the occasion, with very optimistic suggestions. In actual fact, one of many groups discovered an error of their software program throughout a recreation while attempting out the AR system. Tom stated that they’ve acquired plenty of concepts and ideas from the opposite groups for additional developments. This is likely one of the first (if not, the primary) AR system to be trialled throughout the competitors, and first time it has been used within the Normal Platform League. I used to be fortunate sufficient to get a demo from Tom and it positively added a brand new degree to the viewing expertise. Will probably be very fascinating to see how the system evolves.

Mark Subject establishing the MetaQuest3 to make use of the augmented actuality system. Picture credit score: Dr Timothy Wiley.

From the principle soccer space I headed to the RoboCupJunior zone, the place Rui Baptista, an Government Committee member, gave me a tour of the arenas and launched me to a few of the groups which have been utilizing machine studying fashions to help their robots. RoboCupJunior is a contest for varsity kids, and is cut up into three leagues: Soccer, Rescue and OnStage.

I first caught up with 4 groups from the Rescue league. Robots determine “victims” inside re-created catastrophe eventualities, various in complexity from line-following on a flat floor to negotiating paths via obstacles on uneven terrain. There are three totally different strands to the league: 1) Rescue Line, the place robots comply with a black line which leads them to a sufferer, 2) Rescue Maze, the place robots want to research a maze and determine victims, 3) Rescue Simulation, which is a simulated model of the maze competitors.

Group Skollska Knijgia, collaborating within the Rescue Line, used a YOLO v8 neural community to detect victims within the evacuation zone. They skilled the community themselves with about 5000 pictures. Additionally competing within the Rescue Line occasion had been Group Overengeniering2. Additionally they used YOLO v8 neural networks, on this case for 2 components of their system. They used the primary mannequin to detect victims within the evacuation zone and to detect the partitions. Their second mannequin is utilized throughout line following, and permits the robotic to detect when the black line (used for almost all of the duty) adjustments to a silver line, which signifies the doorway of the evacuation zone.

Left: Group Skollska Knijgia. Proper: Group Overengeniering2.

Group Tanorobo! had been collaborating within the maze competitors. Additionally they used a machine studying mannequin for sufferer detection, coaching on 3000 photographs for every sort of sufferer (these are denoted by totally different letters within the maze). Additionally they took photographs of partitions and obstacles, to keep away from mis-classification. Group New Aje had been collaborating within the simulation contest. They used a graphical person interface to coach their machine studying mannequin, and to debug their navigation algorithms. They’ve three totally different algorithms for navigation, with various computational price, which they will swap between relying on the place (and complexity) within the maze through which they’re positioned.

Left: Group Tanorobo! Proper: Group New Aje.

I met two of the groups who had not too long ago introduced within the OnStage occasion. Group Medic’s efficiency was based mostly on a medical situation, with the workforce together with two machine studying components. The primary being voice recognition, for communication with the “affected person” robots, and the second being picture recognition to categorise x-rays. Group Jam Session’s robotic reads in American signal language symbols and makes use of them to play a piano. They used the MediaPipe detection algorithm to seek out totally different factors on the hand, and random forest classifiers to find out which image was being displayed.

Left: Group Medic Bot Proper: Group Jam Session.

Subsequent cease was the humanoid league the place the ultimate match was in progress. The sector was packed to the rafters with crowds desperate to see the motion.
Standing room solely to see the Grownup Measurement Humanoids.

The finals continued with the Center Measurement League, with the house workforce Tech United Eindhoven beating BigHeroX by a convincing 6-1 scoreline. You possibly can watch the livestream of the ultimate day’s motion right here.

The grand finale featured the winners of the Center Measurement League (Tech United Eindhoven) in opposition to 5 RoboCup trustees. The people ran out 5-2 winners, their superior passing and motion an excessive amount of for Tech United.