MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641122839 A) filed by Dr. Safia Naveed. S; and M. Arshiya Mobeen on October 14, 2026, for Method For Dynamic Quantum Memory Allocation Using Reinforcement Learning With A Transformer Policy.
Inventors include Dr. Safia Naveed. S; and M. Arshiya Mobeen.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: A method for dynamic quantum memory allocation in deterministic quantum repeaters using reinforcement learning with a Transformer-based policy is disclosed. The approach models memory management as a Markov Decision Process, where the system state captures memory occupancy, entanglement fidelities, pending requests, and decoherence metrics of stored qubits. A hybrid action space is defined, comprising discrete allocation decisions for activating memory slots, continuous adjustments to storage durations, and binary discard actions to remove degraded qubits. The policy employs a Transformer architecture to learn from temporal sequences of state transitions, addressing the non-Markovian effects of quantum decoherence. A prioritized experience replay buffer stores history windows, with priority based on fidelity decay magnitude and discard events. The agent outputs a mixed probability distribution for both discrete and continuous actions, which are translated into commands for the quantum hardware control plane. A composite reward function balances entanglement completion rate, fidelity quality, and waste penalization due to qubit decoherence. This framework enables adaptive, real-time memory allocation strategies that respect physical constraints, improving end-to-end entanglement generation efficiency in quantum repeater networks.
Disclaimer: Curated by HT Syndication.
