Since the very earliest days of the human race, people have been studying animal behaviours. In those early times, being able to predict animal behaviour gave hunters the advantages required for success. Then, as societies began to develop this gave way, to an extent, to agriculture and early studies, much of it trial and error, enabled farmers to successfully breed and raise livestock to feed an ever growing population. Following the advent of scientific endeavour, more rigorous academic research has taken human understanding of the natural world to much greater depth. In recent years, some of this understanding has been applied to the field of computing, creating the more specialised field of natural computing. In this arena, a considerable amount of research has been undertaken to exploit the analogy between, say, searching a given problem space for an optimal solution and the natural process of foraging for food. Such analogies have led to useful solutions in areas such as numerical optimisation and communication network management, prominent examples being ant colony systems and particle swarm optimisation; however, these solutions often rely on well-defined fitness landscapes that may not always be available. One practical application of natural computing may be to create behaviours for the control of autonomous vehicles that would utilise the findings of ethological research, identifying the natural world behaviours that have evolved over millennia to surmount many of the problems that autonomous vehicles find difficu for example, long range underwater navigation or obstacle avoidance in fast moving environments. This thesis provides an exploratory investigation into the use of natural search strategies for improving the performance of autonomous vehicles operating in a search role. It begins with a survey of related work, including recent developments in autonomous vehicles and a ground breaking study of behaviours observed within the natural world that highlights general cooperative group behaviours, search strategies and communication methods that might be useful within a wider computing context beyond optimisation, where the information may be sparse but new paradigms could be developed that capitalise on research into biological systems that have developed over millennia within the natural world. Following this, using a 2-dimensional model, novel research is reported that explores whether autonomous vehicle search can be enhanced by applying natural search behaviours for a variety of search targets. Having identified useful search behaviours for detecting targets, it then considers scenarios where detection is lost and whether natural strategies for re-detection can improve overall systemic performance in search applications. Analysis of empirical results indicate that search strategies exploiting behaviours found in nature can improve performance over random search and commonly applied systematic searches, such as grids and spirals, across a variety of relative target speeds, from static targets to twice the speed of the searching vehicles, and against various target movement types such as deterministic movement, random walks and other nature inspired movement. It was found that strategies were most successful under similar target-vehicle relationships as were identified in nature. Experiments with target occlusion also reveal that natural reacquisition strategies could improve the probability oftarget redetection.
展开▼
