The coevolution of a team of agents that work together towards a common goal is a complex problem. In this paper, we present the implementation of a learning system that can successfully coevolve a team of four predators to trap a prey in Xpilot-AI. The prey is controlled by a production system and has the ability to avoid crashing into walls and steer away from predators. The predators can be perceived as artificially simulated “lions” and have been given the same capabilities as in real life: they have sensors to sense walls, the prey, and the other predators from a certain distance. They have also been given the ability to communicate with each other. They have to develop searching strategies to find the prey and cooperate to come up with the best strategy to capture it. They are controlled by a production system whose parameters were learned by an evolutionary computation method known as Genetic Algorithms (GA). Similar to real life lions, they learned to encircle the prey prior to capturing it.
Hosanee, Leena Artee, "The Co-Evolution of a Team of Cooperative Autonomous Agents in the Xpilot-AI Environment" (2022). Computer Science Honors Papers. 11.
The views expressed in this paper are solely those of the author.