Evolving Swarm Intelligence for Task Allocation in a Real Time Strategy Game

Abstract

Real time strategy games are complex scenarios where multiple agents must be coordinated in a dynamic, partially observable environment. In this work, we model the coordination of these agents as a task allocation problem, in which specific tasks are given to the agents that are more suited to execute them. We employ a task allocation algorithm based on swarm intelligence and adjust its parameters using a genetic algorithm. To evaluate this approach, we implement this coordination mechanism in the AI of a popular video game: StarCraft: BroodWar. Experiment results show that the genetic algorithm enhances performance of the task allocation algorithm. Besides, performance of the proposed approach in matches against StarCraft's native AI is comparable to that of a tournament-level software-controlled player for StarCraft.

Chain of command in autonomous cooperative agents for battles in real-time strategy games

“…They apply their framework to a StartCraft combat simulator and evaluate it against existing approaches with promising results. 2014; Tavares et al 2017;Yannakakis and Togelius 2015). The complexity of RTS games is enhanced by the fact that player not only has to control the militaristic units but also needs to gather resources to upgrade and develop its military power.…”

“…The complex nature of RTS games is one of the greatest challenges in multi-agent systems. Gathering and using resources, waging war and defending friendly units to achieve an over-arching goal are the basis of decision-making, task allocation and action selection in RTS games (Tavares et al 2014(Tavares et al , 2017.…”

On the Development of Intelligent Agents for MOBA Games

“…Another problem that can be addressed is the development of multiple agents and the interactions between them. Applying approaches of distributing tasks in games, like the presented in [Tavares et al 2014], looks very promising when applied to the MOBA context, as this game genre is strongly oriented to player roles [Yang et al 2014].…”

“…In terms of academic research, there is a rising interest in developing intelligent agents that are capable of playing games that run in real-time; have a dynamic environment; and provide partial information. The similarity of this kind of games to real-world problems is probably one of the main reasons for such research, as the solutions to these problems may be applied to various real-world scenarios [Tavares et al 2014]. This growing interest for game agents, in particular of RTS games, allowed the academia to develop a novel game research scenario, creating competitions of intelligent game agents [Ontanón et al 2013].…”

Research and Implementation of Intelligent Decision Based on a Priori Knowledge and DQN Algorithms in Wargame Environment

“…However, due to the weakness of AI's macro strategic decision-making ability, OpenAI Five failed to beat the professional teams [14,15]. Although some achievements have been made in RTS games, there are still many difficulties in both OpenAI's game AI on Dota and Deepmind's game AI on StarCraft [16][17][18].…”