More Information about Multi Agent challenge

For more information take a look at .


You can download the latest stable multi agent server from here.

You will get an RMASBench folder containing 4 sub-folders (projects):

  • BinaryMaxSum. Library that implements a binary version of MaxSum, including special factors whose messages can be computed more efficiently.
  • Maxsum. Library that implements the standard version of MaxSum.
  • RSLB2. Main tool of the RMASBench platform. This is where most of your work and test will take place, as it allows for an easy interfacing with the robocup rescue simulation platform.
  • BlockadeLoader. Addon simulator that loads blockade definitions from the map file and creates them at the start of the simulation.
  • roborescue Robocup Rescue agent simulation platform.

You need java, ant and maven. you can install them by using following command.

sudo apt-get install ant openjdk-7-jdk maven
#choose openjdk-7-jdk
sudo update-alternatives --config java
sudo update-alternatives --config javac

Make sure your JAVA_HOME is point to the jdk 1.7 .

All this software must be compiled before being able to run the RSLB2 tool. This can be easily achieved by using ant with the proper target for each subfolder:


If everything compiles well (you can ignore warnings, but not errors!), you are now ready to start testing.


Normally, you will run experiments from within the RSLB2/boot folder. Get into that folder and check out the launcher’s options:

cd RSLB2/boot
./ -h

You can now launch an example scenario using any of the included algorithms. When testing, include the -v flag to enable the simulation viewer. For example, you can run the example scenario with the included example experiment configuration by running:

./ -v -c example -m paris -s example-nopolice

This will start a simulation of the example paris scenario without blockades nor police forces. Alternatively, you can run with the police forces and blockades enabled by running the following (notice the -b switch which enables blockades, as well as the change of the scenario file to use):

./ -v -b -c example -m paris -s example-police

Warning: The first time you run a scenario of a map (default is “paris”), the simulator will pre-compute a number of things about the map. This process takes a lot of time (up to one hour depending on your machine), during which you will see the progress on the terminal where you launched the simulation from.

Warning: If error occurred first remove the cache folder. 

When the simulation finishes, you get a results file for each of the algorithms in your experiment. The paths to these results files are displayed in the terminal (right after all agents finish connecting).

Qualified Teams

Agent & Multi-AgentCompetition
Team Name Team Leader
GUC_ArtSapience Dina Helal
Kherad Shirin Reyhanian
MRL Pooya Deldar Gohardani
Poseidon Mahshid Farzan
Ri-one Yosuke Takeuchi
SEU-Jolly Wei NIU
Soshiant Alimohammad Foroutannezhad
Y-Rescue Anderson Tavares
Apollo_Rescue Zhipeng Wei
AUST_KYLIN Zhuchao Chu
CSU_Yunlu Jun Peng
LarvicSaurus Kevin Christian Rodrأ­guez Siu
NAITO-Rescue Nobuhiro Ito
RobotAKUT H. Levent Akin
S.O.S. Hesam Akbari
Infrastructure Competition
Team Name Leader
MRL Pooya Deldar Gohardani
NAITO-Rescue Nobuhiro Ito
Hinomiyagura Masaru Shimizu