Mimicking Canine Olfactory Sensing Using Solid-State Vapour Detectors

One of the most effective techniques employed by the ADF for the detection of explosive materials is the use of explosive detection dogs (EDDs). Despite the advancements in explosive detection instrumentation, the use of EDDs has remained a reliable method to detect explosives due to the dogs’ capacity to discriminate a wide range of vapours, along with the accompanying high sensitivity of their olfactory system. However, the long-term reliance on EDDs has led to increased scrutiny of canine explosive detection capabilities. Notably, there is no robust scientific model to explain the way in which EDDs detect substances.  Furthermore, ADF personnel must maintain a constant human interface with EDDs to monitor their performance, including their response to test samples. These factors present a risk to ADF personnel as there is a requirement for EDD operators to continually monitor canine reactions and behaviour, including in dangerous areas of operation.

In response, there have been several studies conducted which explore the potential for developing a detection system consisting of electronic sensors that perform in a similar fashion to EDDs and their olfactory systems. Although the sensing systems suggested by these studies pose a solution to some of the limitations of EDDs and their scent detection, the detection systems generally lack the sensitivity and selectivity necessary for the conduct of continuous explosive monitoring.

A team of chemistry students conducting Honours research within the Faculty of Science at the Australian Defence Force Academy (ADFA) have been seeking to redress this deficiency.  Specifically, the team is working to develop a detection system, closely resembling that of canine olfactory systems, with a solid-state sensing array and Gas Chromatography Mass Spectroscopy septum connected to a vapour chamber.

The research has set-out to address the question ‘what is the scent environment experienced by an EDD, and how does this relate to its behaviour?’ This question was analysed through the lens of three specific lines of inquiry:  

• Can we use this knowledge to refine training methods, reduce false-negatives and false-positives, and thereby enhance EDD performance?
• Can low-cost solid-state vapour detectors reliably detect explosive materials by mimicking EDD scent logic?
• Can small solid-state vapour detectors be used to complement or improve EDD responses?

Solid-state sensors are small sensing devices utilising semiconductors for their detection. The field of solid-state explosive sensing saw continuous improvement during the period of the ADF’s deployment to Afghanistan due to the need to identify and mitigate the risks posed by adversaries with the capability to manufacture and utilise explosives.

The literature review conducted as part of the ADFA research indicated that the advantages and challenges entailed in operationalising solid-state sensing has been studied extensively. There are also recent studies concerning the implementation of machine learning algorithms on such detection systems. Throughout existing literature on solid-state explosive detection, there are numerous mentions of canine olfactory detection and their efficiency in detection. However, while the utility of canine detection capabilities is universally recognised, there it not yet a complete understanding of the science that underpins their detection methods.

The ADFA research team recognised that modelling the canine detection pathway would involve complex signal processing and pattern recognition facilitated by machine learning. Most existing analysis of solid-state sensory data involves the application of templated methods of sensor data collection which the research team deemed inadequate for their project. In order to produce a classifier and appropriate canine scent profile model, the research team instead sought to generate a bespoke machine data analysis method.

Based on the research conducted to date, the ADFA team has developed an advanced level of understanding concerning the canine olfactory system, canine behaviour pertaining to explosive detection, and has identified the analytical gaps in pre-existing scientific modelling concerning the canine’s detection pathway. The team has also made considerable progress in its research concerning the use of solid-state detection systems, including their advantages and challenges when applied to explosive detection. Further analysis is underway to design an appropriate solid-state sensing array for the detection of energetic systems, including their decomposition products, while employing appropriate machine learning algorithms to process the sensor data.

As advances in solid-state detection devices and machine learning are made, it is more important than ever to address their weaknesses, and to explore their full potential.  For the ADF, generation of a scientific model to explain the way in which EDDs detect substances, supported by a scent profile model and machine data analytical method offer potential improvements to existing EDD training methods.  It also creates opportunities to utilise of EDDs in conjunction with a solid-state sensing array to complement their performance. While much study remains to be done, the research conducted recently at ADFA is a tangible step towards enhancing ADF explosive detection capabilities in future operations.