Search for: All records

A feedforward optical neural network for image recognition is demonstrated through free-space optics in which the nonlinear activation function is provided by a resonant semiconductor saturable absorption mirror (RSAM). Although free-space optics with liquid crystal based spatial light modulator has the potential to allow parallel processing for improved performance, using a fiber-optic electrooptic modulator for parallel to serial conversion greatly simplifies the experimental setup. We show that this nonlinear activation function introduced by RSAM can provide a notable improvement of 8.1% to classification accuracy in comparison to a purely linear network when tested with the MNIST data set for image classification. The impact of noise is also investigated in system implementation. 

The Optical Network Emulation (ONE) engine is a real-time, multi-container software platform designed to model and emulate open optical transport networks with realistic fidelity. This paper introduces an enhanced version of the ONE engine that integrates a distributed implementation of a Gaussian Noise (GN) model for estimating nonlinear interference (NLI) in wavelength division multiplexing (WDM) systems. The inclusion of the GN model enables more realistic emulation of nonlinear signal degradation across diverse link configurations and operating conditions. The enhanced ONE package is then used to document the model’s impact on system performance under varying transmission conditions, including signal launched power and increased spectral loading. With this upgrade, the ONE engine expands its utility for research, development, and education, providing a scalable and flexible environment for testing physical-layer impairments and control strategies in software-defined optical networks. 

A physics-aware simulation platform is proposed for optical neural networks (ONNs), incorporating phase-only modulation and passive free-space propagation. The platform enables end-to-end training under experimentally realistic constraints, with both the phase mask and propagation distance treated as learnable parameters. To facilitate classification, structured 2D output patterns are introduced, where each label corresponds to a fixed spatial light spot. When evaluated with the MNIST dataset, the system achieves 94.6% accuracy using a single phase modulation layer, demonstrating the effectiveness of spatial encoding in physically plausible ONNs. 

With the emergence of IoT applications, 5G, and edge computing, network resource allocation has shifted toward the edge, bringing services closer to the end users. These applications often require communication with the core network for purposes that include cloud storage, compute offloading, 5G-and-Beyond transport communication between centralized unit (CU), distributed unit (DU) and core network, centralized network monitoring and management, etc. As the number of these services increases, efficient and reliable connectivity between the edge and core networks is of the essence. Wavelength Division Multiplexing (WDM) is a well-suited technology for transferring large amounts of data by simultaneously transmitting several wavelength-multiplexed data streams over each single fiber optics link. WDM is the technology of choice in mid-haul and long-haul transmission networks, including edge-to-core networks, to offer increased transport capacity. Optical networks are prone to failures of components such as network fiber links, sites, and transmission ports. A single network element failure alone can cause significant traffic loss due to the disruption of many active data flows. Thus, fault-tolerant and reliable network designs remain a priority. The architecture called “dual-hub and dual-spoke” is often used in metro area networks (MANs). A dual-hub, or in general a multi-hub network, consists of a set of designated destination nodes (hubs) in which the data traffic from all other nodes (the peripherals) should be directed to the hubs. Multiple hubs offer redundant connectivity to and from the core or wide area network (WAN) through geographical diversity. The routing of the connections (also known as lightpaths) between the peripheral node and the hubs has to be carefully computed to maximize path diversity across the edge-to-core network. This means that whenever possible the established redundant lightpaths must not contain a common Shared Risk Link Group (SRLG). An algorithm is proposed to compute the most reliable set of SRLG disjoint shortest paths from any peripheral to all hubs. The proposed algorithm can also be used to evaluate the overall edge-to-core network reliability quantified through a newly introduced figure of merit. 

The Optical Network Emulation (ONE) engine is a software tool that offers students the opportunity to learn how to control and operate open optical (wavelength division multiplexing) transport networks, such as those based on the Open ROADM MSA standards. This paper describes multiple modelling techniques that are implemented in the ONE engine to represent the signal power spectral density at any link/fiber section of the emulated transport network. These techniques make use of polynomial fitting and deconvolution computation methods. 

The Open ROADM ecosystem enables greater flexibility in deploying optical networks through centralized SDN management and intelligent service orchestration. However, since the maintenance signaling is yet to be standardized and implemented within each device, it is difficult to identify the root cause of issues and manage the network effectively when faults propagate in the network. To tackle this problem, this paper presents a proof-of-concept implementation of an alarm correlation and ticketing system for the multi-vendor Open ROADM ecosystem. The proposed system uses a graph-based method to identify the root cause of alarms and generate tickets for network operations teams. The results of our laboratory tests demonstrate the effectiveness of the proposed system in managing alarms in the multi-vendor Open ROADM ecosystem.