US 12,143,832 B2
Neural network circuit remote electrical tilt antenna infrastructure management based on probability of actions
Wenfeng Hu, Täby (SE); Jaeseong Jeong, Solna (SE); Vladimir Verbulskii, Athlone (IE); and Rodrigo Correia, Athlone (IE)
Assigned to Telefonaktiebolaget LM Ericsson (publ), Stockholm (SE)
Appl. No. 17/614,025
Filed by Telefonaktiebolaget LM Ericsson (publ), Stockholm (SE)
PCT Filed Jun. 3, 2019, PCT No. PCT/SE2019/050509
§ 371(c)(1), (2) Date Nov. 24, 2021,
PCT Pub. No. WO2020/246918, PCT Pub. Date Dec. 10, 2020.
Prior Publication US 2022/0248237 A1, Aug. 4, 2022
Int. Cl. H04W 16/24 (2009.01); G06N 3/08 (2023.01); G06N 5/048 (2023.01); H04W 24/08 (2009.01)
CPC H04W 16/24 (2013.01) [G06N 3/08 (2013.01); G06N 5/048 (2013.01); H04W 24/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A remote electrical tilt antenna management computer system comprising:
a network metrics repository that stores live cell performance metrics and stores rule-based data comprising cell performance metrics that were measured during operation of a communication network;
a fuzzy logic circuit having at least one fuzzy inference circuit;
a policy neural network circuit having an input layer having input nodes, a sequence of hidden layers each having a plurality of combining nodes, and an output layer having at least one output node;
at least one processor coupled to the network metric repository, the fuzzy logic circuit, and the policy neural network circuit configured to:
train the policy neural network circuit, when the policy neural network circuit is offline a communication network, to approximate a baseline rule-based policy for controlling a tilt angle of a remote electrical tilt antenna based on the rule-based data stored in the network metrics repository and fuzzy logic data generated by the fuzzy logic circuit applying fuzzy logic to the rule-based data to generate data sets that include a correlated set of a cell performance metric, a tilt angle of a remote electrical tilt antenna, and a reward value to output a probability of actions for a cell under evaluation;
provide to the input nodes of the policy neural network circuit live data received from a live communication network, wherein the live data comprises a live cell performance metric for a cell under evaluation and/or a neighboring cell;
adapt the weights that are used by at least the input nodes of the policy neural network circuit responsive to a policy reward value of an output of the at least one output node of the policy neural network circuit when the policy neural network circuit is in communication with the live communication network; and
control operation of the remote electrical tilt antenna based on output of the at least one output node of the policy neural network circuit, the at least one output node providing the output responsive to processing through the input nodes of the policy neural network circuit at least one live cell performance metric.