A systematic approach to the problem of odour source localisation

Although chemical sensing is far simpler than vision or hearing, navigation in a chemical diffusion field is still not well understood. Biological studies have already demonstrated the use of various search methods (e.g., chemotaxis and biased random walk), but robotics research could provide new ways to investigate principles of olfactory-based search skills (Webb, 2000; Grasso, 2001). In previous studies on odour source localisation, we have tested three biologically inspired search strategies: chemotaxis, biased random walk, and a combination of these methods (Kadar and Virk, 1998; Lytridis et al., 2001). The main objective of the present paper is to demonstrate how simulation and robot experiments could be used conjointly to systematically study these search strategies. Specifically, simulation studies are used to calibrate and test our three strategies in concentric diffusion fields with various noise levels. An experiment with a mobile robot was also conducted to assess these strategies in a real diffusion field. The results of this experiment are similar to those of simulation studies showing that chemotaxis is a more efficient but less robust strategy than biased random walk. Overall, the combined strategy seems to be superior to chemotaxis and biased random walk in both simulation and robot experiment. Christodoulos Lytridis received a BEng degree in electronic and computer engineering, MSc degree in mobile robotics and PhD in Olfactory based navigation for mobile robots from the University of Portsmouth. His research interests include odor-based cooperative navigation, biologically inspired navigational methods, and odor sensors for mobile robots. He is an associate member of the Institution of Electrical Engineering. Endre E. Kadar is a senior lecturer in psychology at the University of Portsmouth. He began his academic career as a theoretical mathematician at the Institute of Hydraulic Planning, working in applied modelling on a variety of problems in engineering. Then he moved to the University of Connecticut to work as a teaching and research assistant and received his PhD in experimental psychology in 1996. His PhD research project has dealt with the problem of ecological foundation of perceptual control of locomotion. After moving to Portsmouth in 1996, his theoretical work provided the basis for a productive collaboration with engineers in robotics. Also, these novel field theoretical ideas have resulted in radically new approaches to exploratory learning and visual control of high speed driving. Application of these approaches has been supported by several ESRC grants. Gurvinder S. Virk graduated with first class honours in electrical and electronic engineering (University of Manchester) and received a PhD in control theory (Imperial College, London). He has followed an academic career working at Sheffield City Polytechnic, Universities of Southampton, Sheffield and Bradford, Portsmouth and is now professor of control and robotics at the University of Leeds. His main research interests include robotics, building and energy management systems, and application of advanced model-based control systems. He has held several grants from national and European sources, as well as industry (the total value of ￡11M). He has more then 200 publications, including four books. He coordinated the European thematic network CLAWAR and has been involved in developing robots for exploring volcanic environments, inspection of highway bridges and dams, and biomedical applications. He is chairman of the ISO TC184/SC2 Advisory Group on Standards for mobile service robots. He is Fellow of the Institution of Engineering and Technology, Fellow of the Chartered Institution of Building Services Engineers, Fellow of the Institution of Applied Mathematics and Its Applications. He is Charted Engineer and a Charted Mathematician. He has been awarded the Freedom of the City of London for his work in promoting Information Technology.