Grant: Canada Research Chairs

The massive push for intelligent cyber-physical systems associated with the Industry 4.0 and the Internet of Things (IoT) revolutions aims at taking better-informed decisions in real-time, based on more complete and readily-acquired sets of data. Better and cheaper flexible hybrid electronics (FHE) sensing devices and circuits designed for gathering and pre-processing the data (so-called edge-computing) will play a pivotal role in these profound transformations. While massive efforts have been invested in developing powerful machine learning (ML) and artificial intelligence tools over the last decade, the sensor designs generally used to gather the information in order to empower (train, test, and validate) those computational models are decades old. As such, they were not designed with the size, weight & power (SWAP) considerations of today’s reality, nor considering the formidable capabilities associated with ML and artificial intelligence. This ambitious program seeks to provide better low-cost and high-performance FHE sensing devices and circuits. Doing so requires significant fundamental research combining new fabrication, integration and deployment strategies. The proposed research will leverage our expertise in order to (Objective #1) provide new and better printable materials and devices and (Objective #2) enable the seamless FHE integration of new inkjet-printed circuit elements with existing technologies through new assembly and post-processing strategies. Accelerating this research for better printed materials and devices remains essential to eventually see those exciting environmentally-greener and cheaper materials spawn viable products. In turn, our team will seek to (Objective #3) demonstrate a significant performance enhancement using recent breakthroughs in deep-learning. Indeed, we will first firmly establish that dramatic accuracy enhancements can be possible using arrays of commercial sensors together with ML capabilities. Then, we will push this concept further by exploring different ML techniques and applying them to our own arrays of low-cost printed sensors to boost their performances. Combining improved FHE platforms with on-chip machine learning capabilities to run inferences locally will eventually lead to better datasets, while alleviating some of the latency and security concerns associated with cloud computing. This will open new possibilities to better exploit the full potential of ML and artificial intelligence.