Communications and Signal Processing
The project includes proposing interference alignment methods and a proof-of-concept demonstration to reduce interference in military communications utilizing efficient spectrum allocation for all terminals. Additionally, flexibility of usage in countries’ borders or conflict zones is aimed in designing next-generation communication systems that can be dynamically configured.
Waveform design deals with effective methods to generate signals and receive them at the receiver side through a channel, meeting the design criteria depending on the demands of users, channel conditions, system, and technology. Waveform design for next generation communication systems including 5G-and-beyond and massive IoT systems is our one of the research interests.
Energy harvesting (EH) from radio frequency (RF) signals is a promising approach that has been regarded as an alternative solution to the power efficiency issue. In EH, nodes can both harvest energy and process information concurrently. As a result, the EH node does not need to have an external source of energy since it uses the harvested energy to forward the received signal to a destination. In the literature, practical EH techniques are classified as power splitting (PS) and time switching (TS) relaying. In PS, the EH node divides the incoming signal power in two parts for EH and information processing (IP) during the whole transmission interval. In TS, a certain fraction of the transmission interval is reserved to EH while the reminder fraction to IP. For both protocols, EH node uses the harvested energy to transmit the received signal to its destination.
Non-orthogonal multiple access (NOMA) is a key-tech for 5G+ mobile communication network as it can provide massive connectivity, low transmission latency, user fairness while improving spectral efficiency and performance at cell edges. In NOMA concept, more than one user with different quality of service levels can access meanwhile the same resource block via power, code or other domains including time (i.e., time division multiple access, TDMA), frequency (i.e., frequency division multiple access, FDMA and orthogonal FDMA, OFDMA) and space (i.e., space division multiple access, SDMA).
Software-defined radios (SDR) are devices that allow to realize dynamic spectrum access through software modifications of SDRs (i.e., adjusting modulation type, coding, filtering, operating on various frequency bandwidths, etc. ). Next generation network protocols (5G and beyond) have been developing through the joint consideration of new technologies (internet-of-things, machine communication etc.).
Cognitive radio (CR) is an intelligent radio that learns from its surrounding environment by analyzing. CR is one of the most promising technologies that aim for efficient spectrum utilization and alleviating the spectrum scarcity problem. Cognitive users or secondary users have to effectively capture the arising spectrum opportunities in time, frequency, and space to transmit their data. Mainly, two aspects characterize the resource allocation for CR networks: 1) primary (licensed) network protection and 2) secondary (unlicensed) network performance enhancement in terms of quality-of-service, throughput, fairness, energy efficiency, etc. CR networks can operate in one of three known operation modes: 1) interweave; 2) overlay; and 3) underlay.
In order to meet the extreme network energy consumption, BS deployment techniques are considered as appropriate solution methods in multi-tier heterogeneous networks (HetNets). Therefore, optimal network planning is considered as a necessary challenge. Grid models are highly idealized models which do not accurately capture the actual BS deployment. So, stochastic geometry models, which handles random spatial patterns, have been commonly used for the deployment of BSs in the HetNet. Instead of hexagonal structure, Poisson Voronoi tessellation (PVT) which is the stochastic geometry scheme model, is suitable for the irregular network topology and much closer to real network structure. The stochastic geometry also allows the examination of the average behavior on many spatial occurrences of a network where network nodes adapt to a given probability distribution. In this way, different types of wireless networks can be modeled, characterized and their behavior can be interpreted. Analysis of energy efficiency methods in k-tier HetNets by using the derived performance metrics (achievable data rate, throughput, coverage probability, etc.) is one of our research topics.
Green communications has been taking great attraction in recent years due to increasing rate of power consumption in rapidly evolving cellular networks. Substantial increase in the number of base stations (BSs) enforces society to focus on green cellular networks for next-generation communication systems. Information and communication technology (ICT) sector will be responsible from 20% of the total energy consumption in 2025 and 14% of global CO2 emissions in 2040. S Considering this concept, significant improvements in energy efficiency at BSs provide not only operational cost savings but also decreased carbon footprint. Thus, it is necessary to utilize more green approaches to meet these concerns.
Time-frequency signal processing is a powerful tool to analyze, process, and interpret nonstationary signals. We developed novel techniques for the analysis and design of high-resolution time-frequency distributions utilizing the fractional Fourier transform. We proposed and analyzed adaptive filtering in fractional Fourier domains in the literature for the first time, and introduced new discrete fractional Fourier transform (DFrFT) definitions. We further worked on the reconstruction of nonuniformly sampled time-limited signals using prolate spheroidal wave functions, and signal recovery in fractional Fourier domains.
Machine learning (ML) is the scientific study of algorithms and statistical models that computer systems use in order to perform a specific task effectively without using explicit instructions, relying on patterns and inference instead. It is seen as a subset of artificial intelligence. Machine learning algorithms build a mathematical model based on sample data, known as “training data", in order to make predictions or decisions without being explicitly programmed to perform the task. Machine learning is closely related to computational statistics, which focuses on making predictions using computers. Neural networks, as the name suggests, are modeled on neurons in the brain. They use artificial intelligence to untangle and break down extremely complex relationships. What sets neural networks apart from other machine-learning algorithms is that they make use of an architecture inspired by the neurons in the brain.