Process Optimisation in Real-Time Domains Using Microcontroller-Based Sensor Devices

Ollencio D'Souza, Subhas Mukhopadhyay, Quan Z. Sheng

Abstract: Life and property are threatened by emergencies that occur in real-time domains with disastrous consequences. It is often the result of a poor awareness of precursors to the event. Such situations cause health issues, fire safety incidents, traffic accidents, negligence and even opportunistic criminality. Managing potential disasters in real-time relies significantly on credible edge intelligence (i.e., knowledge derived from analysed remote event data). Our research presents new methods, techniques and technologies to extract relevant data pre, during and post-event. Early precursor detection and response techniques reduce calamitous occurrences and the threat of emergencies in real-time domains.We demonstrate how microcontroller onboard sensors are "active" and can be tailored to perform effectively in changing, demanding, time-constrained operating environments. They replace passive standard procedural logic with machine learning multifactor associations to improve detection performance "at the edge". We show how "active" sensory capability and relevant data analysis make it possible to extract insights on precursors to manage situational risks more effectively. We discover why and how detecting and recognising a real-time incident's severity is critical to delivering the correct response. The research also uses data from diverse trial location scenarios and data types to improve risk assessment techniques in the targeted real-time domains. Using analysed "site data", we demonstrate how edge intelligence makes it possible to get early warning of events never ordinarily observed before. We have devised a "learning sensor" environment that improves the capability to "predict" and respond to new scenarios that threaten life and property.