AutoRef

This repository contains the source code for the automatic referee for RoboCup SSL tournaments built for the RoboTeam Twente.

Usage

This automatic referee is dependent on the world processing software from RoboTeam Twente. This software package is available as part of our AI docker image. This image can be pulled with:

docker pull roboteamtwente/roboteam:latest

Next one will have to create the actual docker container:

docker run -it --net=host --name roboteam roboteamtwente/roboteam

After creation the container may be started and the observer software can be executed:

docker start roboteam
docker exec roboteam roboteam_world/roboteam_observer

As the project is managed and built using Gradle, any way of using Gradle should work with this project, such as the Gradle integration of your IDE. Explained below is the basic usage guidelines that will require no extra installation steps apart from Java itself.

Make sure you have some distribution of at least Java 17 installed on your machine. On Ubuntu, you can install OpenJDK JRE 17 using the following command:

sudo apt install openjdk-17-jre -y

Finally, you can use Gradle to run the application:

./gradlew run

Program Arguments

The program accepts a few program arguments:

-nogc					turn off the connection to the Game Controller, used for testing without the Game Controller
-cli					run in headless mode
-gpip <value>,--gc-ip <value>		the IP on which the application tries to connect to the Game Controller [default = 127.0.0.1]
-gcp <value>,--gc-port <value>          the port on which the application tries to connect to the Game Controller [default = 10007]
-wip <value>,--world-ip <value>		the IP on which the application tries to connect to the RoboTeam World Observer to [default = 127.0.0.1]
-wp <value>,--world-port <value>	the port on which the application tries to connect to the RoboTeam World Observer to [default = 5558]

When running using gradle, these arguments can be specified in the following way:

./gradlew run --args="-cli -gcp 8642"

Rules

At the moment this automated referee does not implement all the rules that the average automated referee is supposed to validate, in the table below one finds all the rules that are actually being kept track of right now.

Each rule is implemented, however our auto referee cannot reliably make decisions in the Z-plane (height), hence these can be unreliable. Additionally, sometimes objects move too fast, causing us to miss certain fouls.

Rule	Implemented	Implementation	Notes
Attacker double touched ball	[x]	AttackerDoubleTouchedBallValidator.java	Failed the unittest when chipping
Ball left field touch line	[x]	BallLeftFieldTouchLineValidator.java	Fails if the ball leaves mid-air, which is not supposed to happen
Ball left field goal line	[x]	BallLeftFieldGoalLineValidator.java
Aimless kick	[x]	AimlessKickValidator.java	Only applied in division B
Possible goal	[x]	PossibleGoalValidator.java	Goals are detected, but we cannot check whether the ball was kicked or chipped
Defender in defense area	[x]	DefenderInDefenseAreaValidator.java	Failed the unittest when chipping
Attacker touched ball in defense area	[x]	AttackerTouchedBallInDefenseAreaValidator.java	Failed the unittest when chipping
Boundary crossing	[x]	BoundaryCrossingValidator.java
Bot kicked ball too fast	[x]	BotKickedBallTooFastValidator.java
Bot dribbled ball too far	[x]	BotDribbledBallTooFarValidator.java
Penalty kick failed	[x]	PenaltyKickFailedValidator.java
Bot crashing unique	[x]	BotCrashingValidator.java
Bot crashing drawn	[x]	BotCrashingValidator.java
Attacker too close to defense area	[x]	AttackerTooCloseToDefenseAreaValidator.java
Defender too close to kick point	[x]	DefenderTooCloseToKickPointValidator.java
Bot too fast in stop	[x]	BotTooFastInStopValidator.java
Bot interfered placement	[x]	BotInterferedPlacementValidator.java
Placement succeeded	[x]	PlacementSucceededValidator.java