Optical Margin Allocation Using Hybrid Neural Network Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical network design is costly and time-consuming due to the need for reliable information about network topology, which is often unknown or unreliable, making it difficult to create accurate design rules for optical signal transmission systems.
Innovation Solution
A method using a hybrid physical optical model and artificial neural network (ANN) to predict bit error rate (BER) and required optical margin, allowing for reliable optical connection design with minimal network topology knowledge, by training a recurrent neural network (RNN) based on BER excursion parameters and signal quality metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual creation of optical signal transmission system design rules is performed, then reliability of optical transmission is improved, but cost and time consumption increase significantly
Solution Approach 1:
The system performs self-characterization by automatically measuring and learning optical network properties through probe signals, eliminating the need for manual design rule creation. The neural network model trains itself on collected data to predict BER and optimal margins, making the system self-sufficient in characterizing network conditions.
Solution Approach 2:
The patent replaces manual mechanical processes (physical measurements, simulation studies, field tests) with an automated computational system using neural networks. The system substitutes human-in-the-loop design rule creation with algorithm-based predictions that automatically determine optimal optical margins.
2Measurement precision
If simulation studies are performed to obtain reliable information, then accuracy of network characterization is improved, but cost and time consumption increase
Solution Approach 1:
The system performs partial measurements by sending probe signals only on selected connections rather than exhaustively characterizing the entire network. This selective approach gathers sufficient data for training while consuming minimal computational and network resources.
Solution Approach 2:
The system creates a virtual model (copy) of the optical network through neural network training on measured data. This digital twin allows accurate predictions of BER and optimal margins without requiring physical simulations or repeated field measurements.
3Reliability
If field and lab measurements are performed, then reliability of network information is improved, but access difficulty and cost increase
Solution Approach 1:
The system performs self-characterization by automatically collecting network data through integrated probe signals and neural network analysis. This eliminates the need for external field measurements or lab tests, making the process accessible without special equipment or expert intervention.
Solution Approach 2:
The patent introduces probe signals as intermediaries that indirectly characterize network properties without requiring direct physical measurement of difficult-to-access parameters. These probe signals traverse the network and carry information about optical conditions, enabling remote characterization.
4Reliability
If conventional design rules are used, then optical connection reliability is improved, but optical margin allocation efficiency deteriorates
Solution Approach 1:
The system dynamically adjusts the optical margin parameter based on learned network characteristics and specific connection properties. Instead of using fixed conventional margins, the neural network predicts optimal margin values tailored to each connection's actual conditions, improving both reliability and efficiency.
Solution Approach 2:
The patent transitions from static conventional design rules to dynamic, adaptive margin allocation. The system continuously learns from measured data and adjusts margin predictions based on current network conditions, making the allocation process flexible and responsive to changing conditions.
Data Source
AI summary
A system and method for generating, based on optical network topology information, an optical model to represent an optical network; provisioning a new optical connection within the optical network: determining, using the optical model, a first bit error rate (BER) of the new optical connection; determining, using the optical network providing the new optical connection, a second BER of the new optical connection; determining, based on the first and the second BER, a BER excursion parameter of the new optical connection; training a margin allocator based on the BER excursion parameter of the new optical connection, and the first BER of the new optical connection; comparing the first BER of the new connection and a required optical margin to a threshold to determine a reliability of the new optical connection; and allocating, using the margin allocator, the required optical margin for additional optical connections of the optical network.