Soapbox – May
The Shortfalls of Underwater Sensor Network Simulators
By Dr. Mohammad R. Khosravi
In recent years, underwater acoustic sensor networks (UASNs), mostly simulation based, have become a hot research topic. This research covers network protocol design, robotics and autonomous systems, environmental issues, and signal processing.
A key aspect of undersea signal processing is related to sonar systems. Sonar signal processing comprises a large number of signal-processing algorithms to achieve functions such as target detection, localization and classification. One task of sonar is to extract specified features from the acoustic received signals and expound on this information precisely and clearly, without ambiguity for the end-user. The extraction of the specific features and the selection of a proper classifier are essential to the recognition and classification of sonar signals.
For UASN simulation, there are many software tools, including Network Simulator (NS), Network Simulator-v2 (NS2) and Network Simulator-v3 (NS3); NS2 and NS3 are newer versions of NS. In addition to the NS family, OPNET, OMNeT++, QualNet, JSim and many other simulators are open-source tools for network simulation. These are comprehensive platforms for simulating many types of computer and data communication networks, so researchers should use an augmented software package in addition to these platforms to create a proper environment for simulating UASNs. For example, AquaSim can be applied with NS2 to simulate underwater environments.
Although these tools might be helpful for simulating UASN protocols such as routing and MAC (medium access control), there is not yet a complete model as per a real-world underwater network. For instance, in NS2 (along with AquaSim), the model considered for a physical layer (PHY) is very simple. This model uses an ideal inequality to describe a radius range (communication coverage) of messages transmitted from a sender node and a fixed velocity for acoustic waves. In contrast, for some NS3-based packages, the developers have tried to give a more exact PHY model with a nonfixed acoustic wave propagation velocity (in real-world conditions, velocity depends on depth, temperature, etc., whereas some simulators consider an approximate speed of 1,500 m/s in all conditions).
It seems that some parts of a real-world underwater acoustic network cannot easily be captured by simulators, e.g., real communication computations considering all the challenges of a wireless acoustic channel (all losses, channel fading and so on). Complexity of the network protocols design problem and the need for an object-oriented programming environment (because of UASN architecture) for underwater wireless networks add to the challenges for UASN simulators.
A main problem of open-source software tools for network engineers is to use C or C++ along with, for instance, Linux commands and object-oriented programming languages. This might not be a major problem for researchers working in the academic fields of computer science and engineering. However, it is a big and basic obstacle for researchers in electrical engineering who are interested in network simulation and protocol design. Matlab is a typical tool for researchers, and encouraging international developers (academic or industrial) to provide a Matlab-based package for UASNs that doesn’t require skills for different programming languages would open up more resources.
Although there are now some open-source Matlab toolboxes for networks on the internet (Matlab itself is not open source, but additional open-source toolboxes can be supported by Matlab), researchers still require a better, and complete, simulation platform using Matlab to cover all open-system interconnected (OSI) layers under the best modeling. In research areas such as image processing and pattern recognition, this task has been taken up by MathWorks (Matlab’s developer) and independent scientists.
Since Matlab is widely used in academic circles, researchers and scientists must attend to this shortage of network simulation and create new open-source toolboxes. Another possibility is for MathWorks to create such toolboxes for Matlab, similar to existing tools in the latest versions of Matlab. Perhaps a new simulator could have integrated tools for simulating both network protocol design and signal processing. If such options become available, UASN researchers will become more accurate, qualified and efficient in their work.
As a final point, a more comprehensive simulator could enable industrial developers to precisely evaluate the performance of a network before configuration of network devices. This is important because these networks are expensive and key to data communication among sensor nodes for systems in the oil and gas industry, environmental and resource monitoring, marine electronics and ocean energy conversion systems.
Dr. Mohammad R. Khosravi is currently a researcher at the Department of Electrical and Electronic Engineering at Shiraz University of Technology in Shiraz, Iran. Khosravi now serves as a guest editor of Springer Nature Applied Sciences (SNAS), The Journal of Supercomputing and Current Signal Transduction Therapy, and he has served as a guest editor for Current Medical Imaging Reviews.