Real-Time Distributed AI for Networked Cyber-Physical Systems Using IoT Devices and Cloud-Based Data Analytics

Abstract

Digital networks, coupled with physical machine systems, are increasing rapidly. Due to this trend, massive amounts of real-time data are being generated at any given time, necessitating effective ways to handle them in a distributed setting. Here, a different setup is introduced that uses artificial intelligence on local devices and remote computing assistance. It links local sensors, allocates processing power, learns on centralised servers, and modulates behaviour based on perceived outcomes. Incoming sensor reads are processed in simultaneous mode in excess of 12,000; latency is reduced to less than 0.2 seconds. It is accurate to a point of about 93% and almost 90% of the operations are conducted in an efficient manner. Stability is maintained at a maximum of 0.91, without falling below 0.92 in speed consistency, making it more responsive and scalable. As opposed to using one central node, the spread of tasks to more than one node minimizes the communication in the network by over 33%. Compared with traditional methods, the performance remains lower. The machine learning models record 86.8, hybrid models record 89.7, and traditional methods record 74.3, whereas the proposed approach is superior to all others in the main aspects. It can be incorporated into sensor networks and cloud-based real-time analytics, and it also learns under changing and unforeseeable conditions. This can be justified by the fact that decentralized intelligence is more efficient when it comes to swift and smart decision-making in a cyber-physical system where efficiency, stability and scalability play a significant role.