Cross-Layer Energy Optimization for IoT-Enabled Smart Spaces

Abstract

Perpetual IoT systems are essential to many safety and mission-critical applications, e.g. assisted living, healthcare and public safety, which are characterized by continuous monitoring (24/7) and ubiquitous sensing. While IoT-enabled many applications and services, several limitations arise in operating IoT deployments in a resilient manner over time; challenges include the energy cost and constraints. In our research, we aim to handle energy challenges caused by perpetual operations in each level of the system architecture (device, communication, and processing). We use a semantic approach that utilizes context of extracted activities of daily living (ADLs) and indoor space-state (normal, anomaly, and emergency) to drive energy optimized sensor activations. In addition, we are uniquely leveraging features such as: heterogeneity of IoT devices (wearable, ambient, and vision) in terms of: energy cost, energy source (battery-operated and wall-powered IoT devices), processing capability, mobility, communication technologies and transmission protocol (NB-IoT, LTE-M, LoRa, Wi-Fi, 4G/5G, Bluetooth, Zigbee, etc.), processing location (device, edge, could). To validate our approach, we developed an elderly fall detection system using multi-personal and in-situ sensing IoT devices derived from real-world deployments; using our measurements to drive larger simulations. We show that our proposed algorithms such as, Cost-Function-Gradient can achieve greater than 4X reductions in energy dissipation and doubling system-lifetime without loss of sensing accuracy.