Understanding Collective Decision-Making in Natural Swarms

Abstract

In the era of emerging artificially intelligent systems, there is an ongoing struggle to achieve complex behaviours from nature. Swarms of animals exhibit certain emerging behaviour which typically goes beyond the individual level. Nevertheless, each individual acts on its own. Hence, contributes to the emerging behaviour not only by following the swarm but also by considering its own preferences. In this paper, we study the collective decision-making model in artificial swarms and compare it with the natural counterparts. The simulation design in this paper is based on a biological experiment conducted by Miller et al. [1] on swarm of golden shiner fish to achieve a target in multi-target environment. In this paper, we determine, if our model of collective decision-making can interpret the natural behaviours based on the information (personal or social) provided to an individual. In our proposed model, we use the methodologies from the area of multi-criteria decision-making and combine them with the prediction and opinion based approach. Our experiments show that if we add decision-making at individual level based on personal information and refine it with the social information we can achieve a certain collective behaviour as a whole which resembles the existing natural observations. We have proposed a model to achieve consensus in a swarm of fish given a multi-target environment with conflicting criteria. Our results show that personal and social information together play an important role to achieve consensus within swarm.