dorinm17 / roborobo

Braitenberg vehicles tournament

PaintWars Roborobo

Manea Dorin-Mihai & Draghici Mara-Andreea

February, 2024

This project is written in Python and implements Braitenberg vehicles designed using behaviour trees, using the Roborobo simulator. Two teams of eight robots each, that do not communicate with each other, compete to paint as much of the map as possible with their team's color (Red or Blue). One cell of the map can be repainted by the other team that happens to pass by there.

To run: unzip the archive logs.zip and run the go_tournament_eval executable. It will launch a tournament of different arenas and different start positions. Alternatively, run paintwars.py for a single match.
To configurate: paintwars_config.py
Our algorithm is in paintwars_sarmale.py. By default, it competes with the basic implementation paintwars_team_champion.

Each robot has 8 sensors: front, front-left, front-right, left, right, back, back-left, back-right. They distinguish between the different type of obstacles (walls or other robots) and return values ranging from 1 (infinite distance) to 0 (next to). A linear combination of these values and excitatory (+1) or inhibitory (-1) parameters controls the speed of transposition (1 - full speed forward; -1 - full speed backward) and the speed of rotation (1 - 90 deg the right; -1 - 90 deg to the left).

The algorithm prioritizes the avoidance of wall and same team robots, respectively. Next, some robots are designed to follow robots from the opposite team, if detected in proximity, to repaint their occupied cells. All robots explore the map using a genetic algorithm based on mutation (mu=1, lambda=1) to find the best parameters for controlling speed. Escaping measures in case of becoming stuck have been taken into consideration.

Have fun at it!