Distributed Perception Algorithm
In this paper we describe the Distributed Perception Algorithm (DPA) which is partly inspired by the schooling behaviour of ‘golden shiner’ fish (Notemigonus crysoleucas). These fish display a preference for shaded habitat and recent experimental work has
- PDF / 167,895 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 19 Downloads / 181 Views
Abstract. In this paper we describe the Distributed Perception Algorithm (DPA) which is partly inspired by the schooling behaviour of ‘golden shiner’ fish (Notemigonus crysoleucas). These fish display a preference for shaded habitat and recent experimental work has shown that the fish use both individual and distributed perception in navigating their environment. We assess the contribution of each element of the DPA and also benchmark its results against those of canonical PSO.
1
Introduction
The last few decades have seen significant and growing interest in biomimicry or ‘learning from the natural world’, with many disciplines turning to natural phenomena for inspiration as to how to solve particular problems. Examples include the inspiration for engineering designs based on structures and materials found in nature. Another strand of ‘learning from nature’ concerns the development of computational algorithms whose design is inspired by underlying natural processes which implicitly embed computation [5]. Mechanisms of collective intelligence and their application as practical problem-solving tools, has attracted considerable research attention leading to the development of several families of swarm-inspired algorithms including, ant-colony optimisation [6–8], particle swarm optimisation [9,11,12], bacterial foraging [14,15], honey bee algorithms [16] and a developing literature on fish school algorithms. A critical aspect of all of these algorithms is that powerful, emergent, problem-solving occurs as a result of the sharing of information among a population of agents in which individuals only possess local information. A number of previous studies have previously employed a fish school metaphor to develop algorithms for optimisation and clustering. Two of the better-known approaches are Fish School Search (FSS) [1,2] and the Artificial Fish Swarm Algorithm (AFSA) [13]. A practical issue that arises in attempting to develop an algorithm based on the behaviour of fish schools is that we have surprising little hard data on the behavioural mechanisms which underlie their activity. At the level of the individual, agents respond to their own sensory inputs and to their physiological/cognitive states [10]. It is not trivial to disentangle the relative influence of each of these. At group-level, it is often difficult to experimentally observe the mechanics of the movement of animal groups or fish schools. Much previous work developing fish school algorithms has relied on c Springer International Publishing Switzerland 2016 Y. Tan et al. (Eds.): ICSI 2016, Part II, LNCS 9713, pp. 361–369, 2016. DOI: 10.1007/978-3-319-41009-8 39
362
A. Brabazon and W. Cui
high-level observations of fish behaviour rather than on granular empirical data on these behaviours. In this paper, extending initial work in [4], we examine the distributed perception algorithm (DPA). This algorithm draws inspiration from certain behaviours of the fish species ‘golden shiners’. This is a fresh water fish which is native to North America, typically growing to ab
Data Loading...
