The sense of smell is highly important to survival in many animals. Changes in odorous environmental conditions cause most species (ranging from lower invertebrates to mammals) to demonstrate highly adaptive behavioral performances. Natural chemical signals such as food scents or pheromones are particularly unstable and fluctuate in quality, space, and time. Despite this, behavioral responses related to biologically meaningful odor signals can be observed even in complex natural odorous environments where backgrounds may vary enormously, demonstrating that the underlying olfactory neural network is a very dynamic pattern recognition device. We have made significant progress in understanding olfactory receptor transduction mechanisms in biological systems. The chain of events linking the presence of odorant molecules in the air (input) to the generation of a neural spike train (output) has been studied in detail, using biochemistry, molecular biology, and neurophysiology. The chain has been shown to involve odorant binding proteins, odorant degrading enzymes, receptors, G-proteins, effector enzymes, second messengers, and several ionic channels. The key step of the entire transduction process is the interaction of odorant molecules with receptor proteins—each olfactory neuron expressing a single type of receptor. The number of functional receptor types varies depending on the animal species. Odorant molecules interact with several given receptor types with specific affinities. The differential interaction of the population of receptors in the olfactory epithelium finally generates a pattern of active and inactive olfactory receptor neurons, a sensory image, which uniquely codes for the three characteristics of any odorant (or mixture of odorants), its quality, intensity (concentration), and temporal organization. One challenge is to understand how these different aspects are encoded and, on this basis, to reconstruct the global input of the olfactory population to the brain under natural and experimental conditions.

The development of artificial devices for large-scale analysis of chemical environments (such as in the “artificial olfaction” field) suffers from serious drawbacks, at the level of the gas sensors (e.g., nonselectivity, drift) and at the level of the signal processing. Nonselectivity of olfactory receptors is also encountered in biological olfaction, and it might be beneficial to look at the way biological systems process olfactory information, especially because the similarities between early olfactory systems across species imply that nature has found an optimal solution for discriminating odors. Primary olfactory centers in insects and vertebrates show striking similarities both in their cellular organization and in their types of olfactory information coding. Neurophysiological studies have revealed that neuronal circuits in the higher processing areas perform pattern decorrelation, and gain control of population activity, noise reduction, and multiplexing of different information by partial neuronal synchronization.

The book represents a concerted effort by a group of European researchers working together to unravel and translate aspects of olfaction into computationally practical algorithms that may be used in many ways to understand the underlying behavior of the chemical senses in biological systems, as well as being translated into practical applications, such as robotic navigation and systems for uniquely detecting chemical species in a complex background. The material presented is cross-disciplinary and may attract readers with a biological background or an engineering background. Neuromorphic olfaction is developed from conceptual points of view to practical applications in this book. Chapter 1 considers the biological components of vertebrate and invertebrate chemical sensing systems, which are now becoming well understood in terms of their structure and function. Artificial olfaction technology is explained, and multivariate data processing is explained conceptually, while considering biomimetic models of olfaction and how they may be translated into functional engineering devices. Chapter 2 considers the early coding pathways in the biological olfactory system showing how nonspecific receptor populations may have significant advantages in encoding odor intensity as well as odor identity. In Chapter 3 we consider the fact that in the biological system the huge redundancy and the massive convergence of the olfactory receptor neurons to the olfactory bulb are thought to enhance the sensitivity and selectivity of the system. To explore this concept, a modular expandable polymeric chemical sensor array consisting of 16,000 sensors with tens of different types of sensing materials was built and characterized. It is shown that this system has very interesting capabilities in detecting chemicals in a background, as well as discriminating mixtures of chemicals. We turn to a synthetic moth in Chapter 4 showing a neuromorphic approach toward artificial olfaction in robots. This is developed in Chapter 5, where reactive and cognitive search strategies for olfactory robots are discussed. We then consider higher animals, and a computational model of the mammalian olfactory system is discussed and implemented in Chapter 6.

This book represents a considered approach to neuromorphic olfaction that could not have happened without funding. We are grateful to the European Community for supporting our project, Biologically Inspired Computation for Chemical Sensing (NEUROCHEM) (FP7-ICT FET Project 216916).

Krishna C. Persaud

Santiago Marco

Agustín Gutiérrez-Gálvez