5g NR downlink scheduling delay optimization system based on reinforcement learning

Abstract

Based on the strong learning 5G NR downward scheduling delay optimization system, it is based on the state that exists in the actual scenario is not fully observed, and the downlink adjustment process of the base station will be modeling.Learn the algorithm framework to solve.Specifically include: network monitoring module, which is responsible for collecting related inputs of the downward scheduling; resource scheduling module, which is used to simulate the fine granular scheduling process through the simulation device;Some observable status and establish a smart body for different time -scale tasks; the core controller module is used to help the POMDP build module to complete the movement strategy of the smart body of different time -scale tasks in each time slot; the scene adaptation moduleAs the auxiliary module, it provides more scientific and efficient decisions for the core controller, and guides the algorithm to better perform the load balancing in multiple community scheduling scenarios through traffic and space prediction.

Description

technical field [0001] The invention belongs to the field of network systems, in particular to a 5G NR downlink scheduling delay optimization system based on reinforcement learning. Background technique [0002] The development of 5G technology has put forward higher requirements for QoS, such as lower delay, higher data transmission rate, and lower packet loss rate. To meet these challenges, the radio access network should support more advanced waveform technology, larger antennas and more flexible radio resource management. Among them, radio resource management includes transmission power management, mobility management and packet scheduling. As a core component, packet scheduling is responsible for allocating time-domain and frequency-domain resources on the shared channel to users on each TTI. Its purpose is to make trade-offs between four main utilities: capacity (system throughput, spectral efficiency, cell coverage), quality of service (QoS), stability (robustness) ...

AI Technical Summary

Problems solved by technology

[0003] At present, the academic research on air interface delay optimization is mainly focused on the optimization of related algorithms combined with deep learning. A typical example is to use reinforcement learning to update the scheduling strategy in real time through continuous interaction with the network environment to achieve optimal allocation of resources and reduce delay. However, most of the current optimization schemes based on deep learning assume that the state information is completely observable, which is not in line with the actual base station downlink scheduling scenario. Observation information completely from the network environment cannot be obtained; The optimization scheme with the goal of delay

[0004] The technology used in the industry to complete the air interface delay optimization task is mainly the traditional rule-based optimization, which mainly measures a reasonable scheduling function to optimize the delay. Live Update

Images