Artificial Intelligence “at the Edge” for Resource-Constrained Devices Based on SENSIPLUS Technology

Artificial Intelligence (AI) and the proliferation of edge devices have advanced 'AI at the edge', allowing AI models to be deployed directly on devices closer to data sources or end users. Unlike traditional cloud-based approaches that rely on remote data centers, edge AI processes data locally on devices such as smartphones, cameras, and IoT sensors. This local processing reduces latency, lowers bandwidth requirements, improves privacy, and increases reliability. It is valuable for applications with stringent real-time requirements or intermittent connectivity, such as autonomous vehicles, industrial automation, and smart city systems. The rise of edge AI is closely linked to advances in the IoT and Industry 4.0, emphasizing interconnected, autonomous systems. By processing data locally, edge AI reduces network load, allowing IoT devices to perform tasks such as predictive maintenance, real-time monitoring, and autonomous decision-making without cloud delays. However, deploying AI on edge devices is challenging due to limited processing power, memory, and battery life, especially in mobile or resource-constrained environments. Complex AI models require significant computational resources, making optimization techniques such as model compression and quantization essential. Battery-powered devices face additional challenges due to the power requirements of local AI processing, requiring energy-efficient algorithms and specialized hardware. In addition, model updates are required to adapt to dynamic real-world data, creating logistical challenges for consistent updates across devices. Maintaining data quality in an environment of noise is also critical, as poor-quality data can degrade AI performance. The research in this thesis focuses on environmental monitoring and intelligent battery management applications for edge AI. It enables real-time sensor data analysis for rapid environmental response and optimizes battery parameters for energy efficiency in electric vehicles and renewable energy systems. Using Sensichips s.r.l.'s SENSIPLUS sensor platform, these works demonstrate the potential of edge AI in scalable, real-time applications across multiple industries, and highlight the need for high-quality sensors and efficient algorithms in robust edge AI systems. This thesis describes the development and application of Edge AI methods for environmental monitoring and smart battery systems. In environmental monitoring, AI algorithms are used to analyze real-time data from sensors to detect environmental changes. This enabled timely responses, such as triggering alarms for air quality hazards or adjusting system operations. Research also demonstrated the effective detection of pollutants in water for environmental protection. In smart battery systems, the research applied Edge AI to optimize battery performance. This improved battery life prediction and energy consumption. Overall, the research demonstrated that Edge AI can provide robust real-time processing capabilities in environmental and energy-related applications.