MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641107534 A) filed by Dr. Sake Madhu; Mr. Md. Saleem Ahmed; Dr. Neeluri Suresh; Dr. Karnam Sreenu; Mr. B. Maria Joseph; and Mr. V. Narsimha on September 07, 2026, for Predictive Load Balancing In Iot Cloud Platforms Using Multi-Agent Machine Learning Systems.
Inventors include Dr. Sake Madhu; Mr. Md. Saleem Ahmed; Dr. Neeluri Suresh; Dr. Karnam Sreenu; Mr. B. Maria Joseph; and Mr. V. Narsimha.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: ABSTRACT OF THE INVENTION The present invention provides a predictive load balancing system and method for Internet of Things (IoT) cloud platforms that employs a multi-agent machine learning architecture. Conventional reactive load balancers are inadequate for the highly dynamic, spatio-temporal workload patterns generated by large-scale IoT deployments. The invention overcomes these limitations by deploying specialised autonomous agents—Workload Prediction Agents, Resource Allocation Agents, Energy and Carbon Agents, and SLA Monitoring Agents—across edge nodes, regional cloud zones and a global coordination layer. Prediction Agents utilise temporal deep learning models to forecast multi-horizon resource demand. Allocation Agents solve constrained optimisation problems through multi-agent deep reinforcement learning, guided by carbon-intensity and power-cost signals supplied by Energy Agents. SLA Monitoring Agents continuously assess the risk of service-level violations and trigger emergency rebalancing when necessary. A shared distributed knowledge graph serves as a common world model, enabling coherent global decisions from local observations. Hierarchical coordination allows agents to act autonomously under normal conditions while escalating globally significant decisions to a consensus protocol, thereby avoiding a single point of failure. The associated method comprises continuous multi-source telemetry collection, collaborative multi-agent negotiation of placement and scaling plans, proactive execution of migrations and scale actions before predicted peaks materialise, and continuous online as well as federated learning that updates agent policies without exposing raw device data. Experimental evaluation on a simulated smart-city IoT platform demonstrates substantial reductions in peak latency, energy consumption and SLA violation frequency compared with reactive baselines. The invention is applicable to any IoT cloud infrastructure that must balance computational, storage and networking resources under highly variable demand, including smart cities, industrial IoT, connected vehicles and large-scale sensor networks. By shifting from reactive to predictive multi-agent control, the system improves platform efficiency, sustainability and resilience.
Disclaimer: Curated by HT Syndication.
