NUbots have competed in RoboCup every year since 2002.
NUbots have competed in the Four Legged League (2002-2007), the Standard Platform League (2008-2011), the Humanoid KidSize League (2012-2017, 2021-present) and the Humanoid TeenSize League (2018-2019).
NUbots became the Four Legged League world champions in 2006 in Germany. NUbots, in collaboration with the National University of Maynooth from Ireland, won the first Standard Platform League in 2008, as team NUManoid in Suzhou, China.
2024: Eindhoven, The Netherlands
NUbots returned to the Netherlands with eight student team members plus NUbots alumni and 4Ai Systems liaison Trent Houliston, who was reelected to the Humanoid League Technical Committee for a second year. Three team members attended their first RoboCup.
NUbots achieved 4th place with three games won and eight goals scored.
Major hardware changes
- Upgraded ankles and hip pitch and roll to X-Series servos.
- Moved button panel to fixed back panel
- New power solution
- New neck mounts
Major software changes
- YOLO implementation and training of own model
- Hybrid YOLO and Visual Mesh vision object heuristics
- Improved odometry with anchor point method
- Switched field localisation from a particle filter to a non-linear optimiser
- Addition of goals and field line intersections in localisation
- Localisation of other robots on the field
- Torso pitch controller
- Robot-to-robot communication
- Dynamic position decider for defender/striker
- Dynamic bounding box behaviour for different positions, set based on other robots positions, to avoid collisions
- New path planning for positioning to ball, avoiding robots, facing goal
- Kick off and penalty state logic
- NUsight additions: path planning visualisation, goalie/defender/striker visualisation, servo state view, logs view
- RoboCup configuration command line GUI tool
Performance and Results
NUbots placed 14th out of 16 teams in the drop in games.
NUbots placed third in the round robin, missing out on second by goals scored.
| Rank | Team | Points | Played | Won | Drew | Lost | Goals Scored |
|---|---|---|---|---|---|---|---|
| 1 | Ichiro ITS | 9 | 3 | 3 | 0 | 0 | 12 |
| 2 | Hamburg Bit-Bots | 4 | 3 | 1 | 1 | 1 | 5 |
| 3 | NUbots | 4 | 3 | 1 | 1 | 1 | 4 |
| 4 | IT Androids | 0 | 3 | 0 | 1 | 2 | 0 |
NUbots played Bold Hearts in the round of 16 and won three goals to nil. In the quarter-finals, NUbots played TH-MOS and won two goals to nil.
NUbots lost to CIT Brains in the semi-finals, and lost their third place playoff against Barelang FC. The team finished the competition in 4th place out of 16 teams.
2023: Bordeaux, France
RoboCup Bordeaux was the first competition since the pandemic where the team had a full year to prepare without lockdowns keeping the team at home. Eight student members attended, with five attending their first RoboCup competition. NUbots alumni and 4AI Systems liaison Trent Houliston attended as a mentor. During the competition, Trent was voted in by the Humanoid League to the Technical Committee.
Major hardware changes
- OpenCR subcontroller, replacing the CM-740 which is no longer available for purchase
- Upgraded from the NUC7i7BNH to the NUC12WHSi7
- Offset added to the upper leg piece to allow the robot to lift its knees to its chest
- Knee springs added
- Knee servos upgraded to X-Series servos
- Servo covers added to the hip and ankle blocks
Major software changes
- Completely new localisation system using field lines
- Switched from an Unscented Kalman Filter to a Mahony Filter combined with walk engine commands for dead reckoning odometry
- New ball localisation using a Kalman Filter
- Converted from a subsumption-based behaviour system to the Director, a tree-style behaviour system
- With the implementation of the Director, added in more behaviour, such as aligning to the goal, walking to positions in the ready state, and separate striker/defender/goalie behaviour.
- Simplified walk engine based on the Bit-Bots Quintic walk engine
- New kick engine based on the simplified walk engine above
- Newly trained Visual Mesh model fine tuned on the NUbots lab field, which transferred to the Bordeaux field with no further tuning
- New systemconfiguration tool that sets up networking on the robot, allowing for easy switching for the RoboCup fields
- Major upgrades to NUsight, including NBS playback and scrub bar, localisation field lines in the localisation view, zooming in camera view and labelling of cameras.
- Major upgrades to Dockerfile versions, resulting in bug fixes
Performance and Results
NUbots placed 9th in the drop-in games with an average of -3.4 points across five games. A total of 12 teams were in the competition.
NUbots placed second in their first round robin group, with a winning goal against UTRA from Canada.
| Rank | Team | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|---|
| 1 | Rhoban FC | 9 | 3 | 3 | 0 | 0 |
| 2 | NUbots | 4 | 3 | 1 | 1 | 1 |
| 3 | Bold Hearts | 2 | 3 | 0 | 2 | 1 |
| 4 | UTRA | 1 | 3 | 0 | 1 | 2 |
With this result, NUbots progressed to the second round robin directly without having to play in an intermediate game.
NUbots placed third in their second round robin group, drawing against Hamburg Bit-Bots from Germany and losing to a goal by TH-MOS from China.
| Rank | Team | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|---|
| 1 | TH-MOS | 4 | 2 | 1 | 1 | 0 |
| 2 | Hamburg Bit-Bots | 2 | 2 | 0 | 2 | 0 |
| 3 | NUbots | 1 | 2 | 0 | 1 | 1 |
With this unfortunate result, NUbots did not progress to the quarter-finals. The team was close to scoring a goal in both games, but small bugs and issues with the ability of the robots to get up from a fall meant that the team was unable to score. TH-MOS progressed to the semi-finals, where they placed fourth.
Rhoban FC won the competition, with CIT Brains placing second and Barelang FC placing third.
The team made major progress in this competition, with the team scoring a winning goal in a normal game scenario for the first time in seven years. Major software and hardware upgrades made this possible.
2022: Bangkok, Thailand
RoboCup Bangkok was the first in-person RoboCup in three years. All but one team member had never attended an in-person RoboCup competition before. No team members had attended an overseas RoboCup competition. The team faced further issues when the majority of the luggage did not arrive until the end of the first competition day.
Major hardware changes
- Robots upgraded from 3D printed legs to aluminium legs
- New padding and colours
Major software changes
- First time using the Quintic Walk engine in an in-person RoboCup
- Newly trained Visual Mesh with robots and fine-tuned on the Bangkok RoboCup field
- Variety of robots in NUpbr
- New, simpler behaviour using vision balls
- New head behaviour that fixates on the ball and moves in a predefined motion if no ball is found
- New walk path planner that moves the robot to the ball
- Get up fixes, got working on the real robot
Performance and Results
NUbots placed last (11th) in the drop-in games because the team did not have any robots to play with due to missing luggage.
The team had a specially arranged robot inspection at 8.15am on the second competition day and were able to enter the round robin at 11am that day. The team placed second in their round robin group, made up of RO-BIT and ITAndroids. In the intermediate game that followed, the team won in a penalty shootout. This sent the team to the quarter-finals, where they lost to CIT Brains, who won the competition.
NUbots made it to the quarter-finals of the competition. Overall the team played 4 games, won 1, drew 2, and lost 1.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| Round Robins | 2 | 2 | 0 | 2 | 0 |
| Intermediate Game | - | 1 | 1 | 0 | 0 |
| Quarter-Finals | 0 | 1 | 0 | 0 | 1 |
2021/2022 Virtual Season
After the effort of creating the online competition format using Webots in 2021, the League decided to make use of this work by starting the Virtual Season competition. The competition was designed to span over several months, with the first game on the 19th of December 2021, and the final games on the 17th of April 2022.
The Virtual Season included a free three-day workshop following the final games. The workshop had a focus on sim2real and included talks from the 3D Simulation League, team presentations and additional workshops on tools for Webots. The sessions were recorded for future viewing.
Major hardware changes
There was no change to the robot model.
Major software changes
- Webots vision data collection update
- Retrained the Visual Mesh with data from the updated data collection tool
Performance and Results
NUbots were knocked out in the Quarterfinals by 01. RFC Berlin.
| Points | Played | Won | Drew | Lost |
|---|---|---|---|---|
| 4 | 7 | 1 | 1 | 5 |
Brazil Open 2021
The Brazil Open was held online due to COVID-19 with the same setup from RoboCup 2021, allowing the NUbots team to participate.
Major hardware changes
There was no change to the robot model. The competition restricted teams to one robot.
Major software changes
- Fixed the robot freezing issue from RoboCup 2021
- Improved behaviour resulting in the robot more consistently walking to the ball
Performance and Results
The competition involved a round robin, with the winners decided based on game won and goals scored.
| Goals (Own/Against) | Played | Won | Drew | Lost |
|---|---|---|---|---|
| 2 - 21 | 7 | 0 | 5 | 2 |
2021: Online
RoboCup went online in 2021 for all leagues, due to the COVID-19 global pandemic. NUbots participated in the Humanoid KidSize League, which was held using the Webots simulator by Cyberbotics.
Major hardware changes
Because the competition was run virtually in simulation, there were no hardware changes. A robot model with touch sensors on the feet was created for use in the competition.
Major software changes
- Support for Webots
- Major fix/improvement of the Unscented Kalman Filter used for odometry
- Creation of new particle filter for localisation
- Implementation of new walk engine, the Quintic Walk from team Bit-Bots.
- Visual Mesh v2
- NUsight odometry view
Performance and Results
NUbots made it to the quarter-finals of the competition. Overall we played 5 games, won 1, drew 2 and lost 2.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| Round Robins | 5 | 4 | 1 | 2 | 1 |
| Quarter-Finals | 0 | 1 | 0 | 0 | 1 |
2020: Bordeaux, France (cancelled)
The 2020 RoboCup competition was scheduled to occur in the city of Bordeaux in France in June. This was postponed to 2021 due to the COVID-19 global pandemic, but then cancelled early 2021 due to the ongoing pandemic.
2019: Sydney, Australia
RoboCup 2019 was held in Sydney, Australia.
Major hardware changes
- Upgraded from CM730 motor controller and IMU to the CM740.
- Upgraded cameras from FLIR Flea3 to FLIR Blackfly S.
Major software changes
- Changed from lookup table-based vision with a single-class visual mesh to a multi-class visual mesh for vision.
- Added localisation, chart, vision and visual mesh display to NUsight2.
Performance and Results
NUbots made it to the quarter-finals of the competition. Overall we played 13 games, won 5, drew 2, and lost 6.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| Round Robins | 1 | 3 | 0 | 1 | 2 |
| Eliminations for Quarter-Finals (Penalty Shootout) | 3 | 1 | 1 | 0 | 0 |
| Quarter-Finals | 0 | 1 | 0 | 0 | 1 |
| Drop-Ins | -5 | 8 | 4 | 1 | 3 |
2010: Singapore
Major hardware changes
- Upgraded to NAOv3: Intel ATOM 1.6GHz, 1GB Ram
Performance and Results
NUbots were unfortunately knocked out during second round robins by rUNSWift.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| First Round Robin | 6 | 2 | 2 | 0 | 0 |
| Second Round Robin | 2 | 3 | 0 | 2 | 1 |
2009: Graz, Austria
Major hardware changes
- Aldebaran NAOs get an upgrade (NAOv2), dual non-overlapping cameras, so they can see their feet without having to bend over.
Major software changes
- Code purge, removed redundant AIBO Code realised by NUPlatform
- New Git repository, moved away from CVS
- Introduction of threading for NAOs, (SeeThinkThread: 30Hz, SenseMoveThread: 100Hz)
- Blackboard based architecture
- Soft colour classifications: starting using a transition colour, that can belong in different categories of objects, that is assigned an object with context at a later stage in the processing pipeline.
- Advances in Line Detection, with the introduction of corner point (L and T shapes), penalty spot, and center circle detection (ellipse fitting).
- Improvements in Localisation as a direct result of the additional field objects seen.
- Multiple model extended Kalman Filter
- Implemented Goal Keeper behaviour: right dive, left dive, and center
Performance and Results
NUBots were quarter-finalists in 2009.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| Round Robins | 13 | 5 | 4 | 1 | 0 |
| Quarter-Finals | 0:1 | 1 | 0 | 0 | 1 |
2008: Suzhou, China
Major hardware changes
- First year with the Aldebaran NAOs (x86 AMD GEODE 500MHz CPU, 256 MB SDRAM / 2 GB flash memory)
- Low-powered walk optimisation
Major Software changes
- Updated AIBO 2007 code base to interface with NAOQI (using Aldebarans middleware for communicating with NAOs low-level hardware).
Performance and Results
NUManoids (The University of Newcastle (NUbots) partnered with National University of Maynooth from Ireland) - First place, Standard Platform League.
| Round | Points | Played | Won | Drew | Lost |
|---|---|---|---|---|---|
| Round Robins | 12 | 4 | 1* | 3 | 0 |
| Quarter-Finals | 2:0 | 1 | 1 | 0 | 0 |
| Semi-Finals | 1:0 | 1 | 1 | 0 | 0 |
| Grand-Final (Time to Score Goal) | 1:0 | 1 | 1 | 0 | 0 |
- The penalty shootout was judged based on the distance of the ball to the goal after three minutes.
- Reference: https://spl.robocup.org/history/nao-results-2008/
