Smart City Threat Detection Using Integrated Chemical Sensing and Machine Learning

This project aims at expending the AirU smart cities air quality monitors into a highly sensitive and adaptive, chemical and explosive threat identification embedded system, which can be deployed on mobile or fixed platforms. A chemical sensing network requires (1) its sensing elements and read-out circuitry to be highly selective and sensitive to different chemicals, and (2) the raw data to be processed through highly adaptive algorithms, accounting for environmental conditions (humidity, temperature, pollution, wind, etc.) to perform educated background removal and obtain low-false alarm rates.

The chemical sensing will be performed by an array of high-selectivity and sensitivity (at ppt level) nanofiber chemiresistive sensors. Current sensor technologies applicable to chemical threat detection generally fall into four categories: spectrometric, optical, chemiresistive, and microelectromechanical systems (MEMS). Among them, chemiresistive organic nanofibers are the only technology that can be implemented at low cost and in a portable form factor without substantial loss of sensitivity or selectivity compared to lab-based instruments. The sensing organic nanofibers used in this project are developed and will be fulfilled by Vaporsens, Inc.

The proposed sensor elements will be interfaced with the AirU, an existing Smart City environment sensor platform developed by Prof. Gaillardon, and capable of measuring environmental factors (temperature, humidity, etc.) and species that can indicate the trajectory of a chemical threat (i.e., CO2, PM). The proposed solution is illustrated in Fig. 3. A high-sensitivity (at pA level) ASIC interface will acquire the sensing data and convert to 8-bit digital formats. The digital sensor data will then be processed on-board to correct the signatures that may be obscured by the local atmospheric conditions.

Fig.: Overarching view of the technical developments: (a) 10 nanofiber sensors will be integrated on a borosilicate glass substrate along with their data acquisition ASIC, and packaged in a housing for optimal air flow; (b) The AirU smart city monitors will be upgraded to run an efficient machine-learning-based background-removal algorithm and communicate with the SIGMA+ network.

This research effort is funded by DARPA and the Department of Interior agreement number D19AP00028.