Swarm intelligence for autonomous cooperative agents in battles for real-time strategy games

Abstract

Abstract-this paper investigates the use the swarm intelligence of honey bees to create groups of co-operative AI for an RTS game in order to create and re-enact battle simulations. The behaviour of the agents are based on the foraging and defensive behaviours of honey bees, adapted to a human environment. The groups consist of multiple model-based reflex agents, with individual blackboards for working memory, with a colony level blackboard to mimic the foraging patterns. An agent architecture and environment is proposed that allows for creation of autonomous cooperative agents. The behaviour of agents is then evaluated and their intelligence is tested using an adaptation of Anytime Universal Intelligence Test.

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

“…This section provides a brief description of the evaluation method as presented by Daylamani-Zad et al (2017). In this approach for evaluating the intelligence of a multi-agent system, each agent has its own specific role and each role would have its own specific set of tasks and states that it is supposed to achieve.…”

“…The evaluation method used for Unit Intelligence is an expansion of the previous evaluation used by Daylamani-Zad et al (2017) which was inspired by Chmait et al (2016b) and their approach to anytime universal intelligence test (Hernández-Orallo and Dowe 2010). For this purpose, the environment is created based on a K environment (Chmait et al 2016a;Insa-Cabrera et al 2012).…”

“…The goal is to recreate autonomous AI that can be used to re-enact battles in a simulation format, which could also be used in history or military training. This work builds on previous implementation of Cooperative agents (Daylamani-Zad et al 2017) and addresses the challenge of command structure, allowing for new outcomes such as retreating and surrender.…”

University of Greenwich

“…They produce a product as a proof-of-concept in order to evaluate/validate the hypothesis. This has led to final year projects being presented at industry-led conferences (Sheehan 2017) and published in academic proceedings (Daylamani-Zad, Graham & Paraskevopoulos 2017).…”

An Adaptive Moth Flame Optimization Algorithm with Historical Flame Archive Strategy and its Applications

“…Swarm intelligence algorithm is a kind of nature-inspired stochastic algorithm for searching optimal solution of the nonlinear, non-differentiable and non-separable complex problem by simulating the foraging behaviors and biological habits of animals, and has received increasing attention in last several decades [1][2][3][4][5][6][7][8][9][10]. The classical swarm intelligence algorithms have Particle Swarm Optimization (PSO) [11], Artificial Bee Colony (ABC) [12] and Ant Colony Optimization (ACO) [13], etc.…”