The pyGOMAS Multi-Agent Strategy Competition

PyGOMAS (Python Game Oriented Multi-Agent System) is a framework for developing and simulating multi-agent systems (MAS) in a game-like environment. The pyGOMAS Multi-Agent Strategy Competition, leveraging this framework, invites teams to program AI agents for strategic challenges within a virtual battlefield. In the competition, the teams’ agents will demonstrate their capabilities in performing tasks like flag capture or defense. Adhering to strict rules for fair play and agent autonomy, this event not only tests AI strategy and teamwork but also promotes innovation and learning in the field of MAS.

Free software

Get the code of pygomas for free

BDI Agents

Use a BDI development envvironment to win a Capture the Flag competition!

BaseD on SPADE

pygomas uses the well-known Python framework for developing multiagent systems: SPADE

SUbmit your participation

Create your own BDI soldiiers and fight for winning the flag

Submit your participation!

Important dates

  • Call for contributions: February 22, 2024
  • Dedline for submissions of contributions: March 27, 2024
  • Notification to competition finalists: April 10, 2024
  • AAMAS’24 competition: May 8-9, 2024

Competition Rules

The competition will be carried out by a set of 1 vs 1 contests between the different participating players, according to an schedule that will be published by the organizers. Each player will have to provide in advance the pyGOMAS files defining a valid strategy for a team to play the game.

Each contest between two players will be carried out by playing two matches, following these competition rules:

  • Duration: 5 minutes
  • Number of agents per team: 10 agents.
  • Map: previously unknown
  • Distribution: each team will play as Allied in one match, and as Axis in the other.

After each match, a Result Value (RV) will be computed and assigned to each player. The RV of a match is calculated as the sum of the following values:

  • 100 points for the team winning the match (the Allied team wins if captures and carries the flag to its base before the time is out; the Axis team wins in any other case).
  • 5 points for each surviving agent in the team at the end of the match.
  • 0.1 * the average health of the surviving agents at the end of the match.

After the two matches, each player gets the sum of their RVs, and the player with the highest value will win the contest. If there is a tie (both players get exacty the same accummulated RV), a third match will be played, where Allied and Axis teams will be randomly assigned.

No Alteration of Agent States

Participants are not allowed to directly alter any agent’s state (life, ammunition, etc.) in the game environment. All agent state changes must result from in-game interactions and decisions.

Communication Restrictions

Teams must only use in-game messaging systems for communication between their agents.

Direct data interchange through external means is prohibited.

Communication with the opposing team’s agents is strictly forbidden.

Teams must rely solely on their strategies and the agents’ ability to interpret the game environment.

Code Integrity

Tampering with the pyGOMAS framework or other teams’ code is strictly prohibited. All code must be original or appropriately credited if external libraries or snippets are used.

Fair Play

Teams must adhere to the spirit of fair competition. Any form of cheating, exploiting bugs, or unsportsmanlike conduct will result in disqualification.

Code examples

Files to execute your first pyGOMAS game

Here you can find some example files to build your team, and execute your first pyGOMAS game:
– JSON file for loading the agents in the game (after Manager and Service agent are created, and connected to an XMPP Server. These 3 names should be personalized in this JSON file).
– ASL files of the 3 different basic roles being played by pyGOMAS troop agents: soldier (bdisoldier.asl), medic (bdimedic.asl) and FieldOps (bdifieldop.asl).

Our Team

We are five members from the Valencian Research Institute for Artificial Intelligence (VRAIN)

Javier Palanca

VRAIN

Vicente Julian

VRAIN

Andrés Terrasa

VRAIN

Carlos Carrascosa

VRAIN

Vicent Botti

VRAIN

HELP & SUPPORT

Any questions about the competition or pyGOMAS can be sent to:
jpalanca AT dsic.upv.es