US 12,470,435 B2
Deep learning method and system for spectrum sharing of partially overlapping channels
Lu Wang, Shenzhen (CN); Ruifeng Huang, Shenzhen (CN); and Kaishun Wu, Shenzhen (CN)
Assigned to Shenzhen University, Shenzhen (CN)
Appl. No. 18/281,418
Filed by Shenzhen University, Shenzhen (CN)
PCT Filed Mar. 26, 2021, PCT No. PCT/CN2021/083271
§ 371(c)(1), (2) Date Sep. 11, 2023,
PCT Pub. No. WO2022/198634, PCT Pub. Date Sep. 29, 2022.
Prior Publication US 2024/0121136 A1, Apr. 11, 2024
Int. Cl. H04W 28/082 (2023.01); H04L 25/02 (2006.01); H04L 5/00 (2006.01)
CPC H04L 25/0254 (2013.01) [H04W 28/082 (2023.05); H04L 5/0057 (2013.01)] 9 Claims
OG exemplary drawing
 
1. A deep learning method for spectrum sharing of partially overlapping channels, comprising the following steps:
in response to a received user transmission request, inputting, by a base station, channel state information CSI of a plurality of historical time slices into a trained channel prediction convolutional neural network model and outputting predicted channel state information CSI of a next time slice; and
inputting the channel state information CSI of the next time slice into a reinforcement learning model and obtaining a channel allocation strategy of each user equipment in a collision domain of the base station to realize a maximum throughput of simultaneous communication of each user equipment, wherein the reinforcement learning model is obtained by training by taking bandwidth efficiency performance as a reward;
wherein the inputting the channel state information CSI of the next time slice into a reinforcement learning model and obtaining a channel allocation strategy of each user equipment in a collision domain of the base station to realize a maximum throughput of simultaneous communication of each user equipment comprises the following steps:
assuming that there are n pieces of user equipment UE in the collision domain of the base station, Stotal represents a total number of blocks in a channel, CSIi is a channel state, si and Pi represent allocated blocks and overlaps of UEi, respectively, ri represents an achievable data rate of UEi under the overlap Pi, the goal is to maximize the overall throughput as Ui=sumi−1nri, and the reinforcement learning model is used to find out an optimal allocation strategy under different channel states, wherein the channel state information CSI of each user equipment is input to the reinforcement learning model as an input, and an agent action is a partially overlapping channel weight allocated to each user equipment, and an action space includes all available allocations of partially overlapping channel weights.