MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202621082014 A) filed by Symbiosis International Deemed University on July 02, 2026, for Shielded Hybrid Reinforcement Learning System For Energy-Efficient And Reliable Virtual Machine Migration In Clouds.
Inventors include Himanshu Mahakalkar; and Dr. Bhupesh Kumar Dewangan.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: ABSTRACT SHIELDED HYBRID REINFORCEMENT LEARNING SYSTEM FOR ENERGY-EFFICIENT AND RELIABLE VIRTUAL MACHINE MIGRATION IN CLOUDS The present invention discloses a shielded hybrid reinforcement learning system (100) and method for energy-efficient and reliable virtual machine migration in cloud data centers. The system (100) comprises a four- layer hierarchical architecture including a data observation layer (110) for telemetry collection and state vector generation, an intelligence layer (130) implementing multi-agent deep reinforcement learning with Proximal Policy Optimization policy networks under a centralized training distributed execution paradigm, a constraint-satisfaction safety shield (140) validating migration proposals against resource capacity, SLA threshold, and stability rules before execution, and an execution layer (150) for hypervisor-interfaced migration execution and reward-based policy refinement. A digital twin simulation environment (160) enables offline pretraining on historical workload traces for warm-started deployment. The system achieves 12.2 percent energy savings over static heuristics while maintaining 96 percent reduction in SLA violations compared to unshielded RL systems, enabling autonomous, safe, and scalable cloud resource management. [
Disclaimer: Curated by HT Syndication.
