Forecasting Evolution of Clusters in Game Agents with Hebbian Learning

TL;DR

This paper introduces a hybrid AI model combining Hebbian learning and LSTM to forecast cluster evolution in multi-agent systems like StarCraft II, improving efficiency and predictive accuracy of agent group dynamics.

Contribution

It presents a novel hybrid model that integrates unsupervised Hebbian clustering with LSTM-based forecasting for multi-agent system dynamics.

Findings

Successfully predicts complex cluster movements in StarCraft II

Achieves lower inference time than traditional clustering methods

Demonstrates effective long-term cluster evolution forecasting

Abstract

Large multi-agent systems such as real-time strategy games are often driven by collective behavior of agents. For example, in StarCraft II, human players group spatially near agents into a team and control the team to defeat opponents. In this light, clustering the agents in the game has been used for various purposes such as the efficient control of the agents in multi-agent reinforcement learning and game analytic tools for the game users. However, despite the useful information provided by clustering, learning the dynamics of multi-agent systems at a cluster level has been rarely studied yet. In this paper, we present a hybrid AI model that couples unsupervised and self-supervised learning to forecast evolution of the clusters in StarCraft II. We develop an unsupervised Hebbian learning method in a set-to-cluster module to efficiently create a variable number of the clusters with lower inference time complexity than K-means clustering. Also, a long short-term memory based prediction module is designed to recursively forecast state vectors generated by the set-to-cluster module to define cluster configuration. We experimentally demonstrate the proposed model successfully predicts complex movement of the clusters in the game.