Search for: All records

Abstract Neuromorphic computing, exemplified by breakthroughs in machine vision through concepts like address-event representation and send-on-delta sampling, has revolutionised sensor technology, enabling low-latency and high dynamic range perception with minimal bandwidth. While these advancements are prominent in vision and auditory perception, their potential in machine olfaction remains under-explored, particularly in the context of fast sensing. Here, we outline the perspectives for neuromorphic principles in machine olfaction. Considering the physical characteristics of turbulent odour environments, we argue that event-driven signal processing is optimally suited to the inherent properties of olfactory signals. We highlight the lack of bandwidth limitation due to turbulent dispersal processes, the characteristic temporal and chemical sparsity, as well as the high information density of the odour landscape. Further, we critically review and discuss the literature on neuromorphic olfaction; particularly focusing on neuromorphic principles such as event generation algorithms, information encoding mechanisms, event processing schemes (spiking neural networks), and learning. We discuss that the application of neuromorphic principles may significantly enhance response time and task performance in robotic olfaction, enabling autonomous systems to perform complex tasks in turbulent environments—such as environmental monitoring, odour guided search and rescue operations, and hazard detection. 

Odours released by objects in natural environments can contain information about their spatial locations. In particular, the correlation of odour concentration timeseries produced by two spatially separated sources contains information about the distance between the sources. For example, mice are able to distinguish correlated and anti-correlated odour fluctuations at frequencies up to 40 Hz, while insect olfactory receptor neurons can resolve fluctuations exceeding 100 Hz. Can this high-frequency acuity support odour source localization? Here we answer this question by quantifying the spatial information about source separation contained in the spectral constituents of correlations. We used computational fluid dynamics simulations of multisource plumes in two-dimensional chaotic flow environments to generate temporally complex, covarying odour concentration fields. By relating the correlation of these fields to the spectral decompositions of the associated odour concentration timeseries, and making simplifying assumptions about the statistics of these decompositions, we derived analytic expressions for the Fisher information contained in the spectral components of the correlations about source separation. We computed the Fisher information for a broad range of frequencies and source separations for three different source arrangements and found that high frequencies were more informative than low frequencies when sources were close relative to the sizes of the large eddies in the flow. We observed a qualitatively similar effect in an independent set of simulations with different geometry, but not for surrogate data with a similar power spectrum to our simulations but in which all frequencies werea prioriequally informative. Our work suggests that the high-frequency acuity of olfactory systems may support high-resolution spatial localization of odour sources. We also provide a model of the distribution of the spectral components of correlations that is accurate over a broad range of frequencies and source separations. More broadly, our work establishes an approach for the quantification of the spatial information in odour concentration timeseries. 

Mice navigate an odor plume with a complex spatiotemporal structure in the dark to find the source of odorants. This article describes a protocol to monitor behavior and record Ca2+ transients in dorsal CA1 stratum pyramidale neurons in hippocampus (dCA1) in mice navigating an odor plume in a 50 cm x 50 cm x 25 cm odor arena. An epifluorescence miniscope focused through a GRIN lens imaged Ca2+ transients in dCA1 neurons expressing the calcium sensor GCaMP6f in Thy1-GCaMP6f mice. The paper describes the behavioral protocol to train the mice to perform this odor plume navigation task in an automated odor arena. The methods include a step-by-step procedure for the surgery for GRIN lens implantation and baseplate placement for imaging GCaMP6f in CA1. The article provides information on real-time tracking of the mouse position to automate the start of the trials and delivery of a sugar water reward. In addition, the protocol includes information on using of an interface board to synchronize metadata describing the automation of the odor navigation task and frame times for the miniscope and a digital camera tracking mouse position. Moreover, the methods delineate the pipeline used to process GCaMP6f fluorescence movies by motion correction using NorMCorre followed by identification of regions of interest with EXTRACT. Finally, the paper describes an artificial neural network approach to decode spatial paths from CA1 neural ensemble activity to predict mouse navigation of the odor plume. 

In order to survive, animals often need to navigate a complex odor landscape where odors can exist in airborne plumes. Several odor plume properties change with distance from the odor source, providing potential navigational cues to searching animals. Here, we focus on odor intermittency, a temporal odor plume property that measures the fraction of time odor is above a threshold at a given point within the plume and decreases with increasing distance from the odor source. We sought to determine if mice can use changes in intermittency to locate an odor source. To do so, we trained mice on an intermittency discrimination task. We establish that mice can discriminate odor plume samples of low and high intermittency and that the neural responses in the olfactory bulb can account for task performance and support intermittency encoding. Modulation of sniffing, a behavioral parameter that is highly dynamic during odor-guided navigation, affects both behavioral outcome on the intermittency discrimination task and neural representation of intermittency. Together, this work demonstrates that intermittency is an odor plume property that can inform olfactory search and more broadly supports the notion that mammalian odor-based navigation can be guided by temporal odor plume properties. 

This Perspective highlights the shift from the classic picture of olfaction as slow and static to a view in which dynamics play a critical role at many levels of sensing and behavior. Olfaction is now increasingly seen as a “wide-bandwidth temporal sense” (Ackels et al., 2021; Nagel et al., 2015). A parallel transition is occurring in odor-guided robot navigation, where it has been discovered that sensors can access temporal cues useful for navigation (Schmuker et al., 2016). We are only beginning to understand the implications of this paradigm-shift on our view of olfactory and olfacto-motor circuits. Below we review insights into the information encoded in turbulent odor plumes and shine light on how animals could access this information. We suggest that a key challenge for olfactory neuroscience is to re-interpret work based on static stimuli in the context of natural odor dynamics and actively exploring animals.