To address growing wireless data processing demands in telecommunications and radar sensors, heterogeneous multiprocessor systems-on-chip (MPSoC) integrating programmable processors and hardware accelerators are increasingly being utilized. As the number of multiple-input multipleoutput (MIMO) elements to process data increases in telecommunications and radar system on chips (SoCs), computational complexity and area of these SoCs also grows. Use of MIMO technology also increases the volume of data transfer across the interconnects and also data storage requirements in on-chip memories of the MPSoCs. Data compression is an important way to reduce memory sizes, data transfer time, and communication bandwidth requirements in wireless data processing SoCs. In this article, we focus on efficient memory and bandwidth utilization data compression techniques for wireless data processing SoCs with applications in telecommunications and automotive radars. To this end, our contribution is twofold: First, we present a comprehensive literature review of both lossy and lossless compression techniques playing a critical role in reducing band-width utilization across SoC interconnects and minimizing onchip memory usage. We provide a classification of the use of compression techniques utilized for data compression, and also propose a comprehensive taxonomy of compression techniques for wireless data processing SoCs. We make the case that, although automotive radar SoCs currently utilize data compression techniques to save on-chip memory sizes, compression during in-phase and quadrature (IQ) data transfers from the front-end to the baseband is now essential for MIMO radars. This is due to the large number of receive processing chains being incorporated in these radars to improve their resolution to meet the stringent requirements of autonomous driving. Our second contribution is to design data compression for automotive radar IQ data. Using real radar data collected by us, via the Texas Instruments (TI) AWR1642 radar SoC, we present radar IQ data analysis. We also present two radar data compression techniques for radar data transfer and show that block floating point (BFP) compression performs better than block scaling compression (BSC). Our results show that the automotive radar IQ data can be compressed 50% without degrading the radar detection performance.

The rapid adoption of electric vehicles (EVs) poses significant challenges for large-scale EV charging stations in terms of efficiently addressing dynamic charging demands and proactively managing grid load. This paper proposes a cost-efficient framework that extends the$M/M/s$queuing model to a time-varying$M(t)/M/s(t)$system, enabling adaptive charging session management based on uncertainty-aware predicted EV arrivals. A Bayesian Neural Network (BNN)-based model generates mean and uncertainty-aware upper bound (UB) arrival predictions, while a novel cost function, with linear and non-linear waiting costs and a tolerable waiting time, balances grid capacity reservation costs with user waiting time costs. The simulation results demonstrate that UB EV arrival predictions ensure robust performance under uncertainties. The non-linear cost model with UB prediction effectively reduces maximum waiting times by approximately 30% compared to the linear model, improving user satisfaction. Key findings highlight that incorporating tolerable waiting times of 10 to 20 minutes reduces total cost by approximately 12% to 25% relative to the no-threshold approach. Moreover, the adaptive demand estimation strategies outperform static approaches by dynamically adjusting to varying EV arrivals. The proposed framework offers a scalable and practical solution for managing operational costs and user expectations in future EV charging infrastructures.

To enable widespread use of electric vehicles (EVs), large-scale public charging stations with fast chargers are being planned in places such as shopping malls and office car parks. Operators of public charging stations need to utilize EV scheduling algorithms that can satisfy charging demands with a minimum number of simultaneous charging sessions. In this paper, we propose EV charging scheduling algorithms that meet the charging demands and deadlines of EV users while minimizing the number of simultaneous charging sessions. An uncertainty-aware deep learning (DL) framework is also used to predict EV arrivals at a charging station. The predicted EV arrivals in turn are used to help the charging operator estimate how many charging sessions to order from the grid. Our DL model not only predicts the mean EV arrival rates but also the upper limits of EV arrivals, which enhances robustness against uncertainty in EV arrivals and helps estimate the maximum charging demand for a given interval. Moreover, to overcome the challenge of insufficient EV charging data for DL models, we construct a synthetic data model that takes into account multiple factors influencing EV arrivals, such as weather, events, weekdays, and weekends. Both online and offline approaches in the design of EV scheduling algorithms are utilized. The performances of the proposed algorithms are evaluated in terms of active charging sessions used to serve EV users. We also compare their performance with a baseline algorithm which is an offline optimal algorithm based on a mixed integer linear problem formulation.

Proactive and intelligent management of network resource utilization (RU) using deep learning (DL) can significantly improve the efficiency and performance of the next generation of wireless networks. However, variations in wireless RU are often affected by uncertain events and change points due to the deviations of real data distribution from that of the original training data. Such deviations, which are known as dataset drifts, can subsequently lead to a shift in the corresponding decision boundary degrading the DL model prediction performance. To address these challenges, we present hardware-accelerated real-time radio frequency (RF) analytics and drift-awareness modules for robust DL predictions. We have prototyped the proposed design on a Zynq-7000 System-on-Chip that contains an FPGA and an embedded ARM processor. We have used Xilinx Vivado design suite for synthesis and analysis of the HDL design for the proposed solution. To detect dataset drifts, the proposed solution adopts a distance-based technique on FPGA to quantify in real-time the change between the prediction distribution obtained from DL predictions and data distribution of input streaming samples. Using various performance metrics, we have extensively evaluated the performance of the proposed solution and shown that it can significantly improve the DL model robustness in the presence of dataset drifts.

Deep learning models usually assume that training dataset and target data have the same distribution. If this is not the case, model mismatch causes performance degradation when the model is used with the real data. With radio frequency (RF) measurements from real data traffic, the exact distribution of the measurements is unknown in many cases and model mismatch is unavoidable. This is known as concept drift, or model mis- specification in deep learning, which we are interested in for cognitive radio dynamic spectrum access predictions. In this paper, we present three concept drift detection methods and their corresponding very large scale integration (VLSI) circuits. The circuits are mapped on a Xilinx Virtex-7 field-programmable gate array (FPGA) and the resource utilization results are provided.

This paper addresses the following question which is of interest in designing efficient exclusive-use spectrum licenses sold through spectrum auctions. Given a system model in which customer demand, revenue, and bids of wireless operators are characterized by stochastic processes and an operator is interested in joining the market only if its expected revenue is above a threshold and the lease duration is below a threshold, what is the optimal lease duration which maximizes the net customer demand served by the wireless operators? Increasing or decreasing lease duration has many competing effects; while shorter lease duration may increase the efficiency of spectrum allocation, longer lease duration may increase market competition by incentivizing more operators to enter the market. We formulate this problem as a two-stage Stackelberg game consisting of the regulator and the wireless operators and design efficient algorithms to find the Stackelberg equilibrium of the entire game. These algorithms can also be used to find the Stackelberg equilibrium under some generalizations of our model. Using these algorithms, we obtain important numerical results and insights that characterize how the optimal lease duration varies with respect to market parameters in order to maximize the spectrum utilization. A few of our numerical results are non-intuitive as they suggest that increasing market competition may not necessarily improve spectrum utilization. To the best of our knowledge, this paper presents the first mathematical approach to optimize the lease duration of spectrum licenses.

Learning the channel occupancy patterns to reuse the underutilised spectrum frequencies without interfering with the incumbent is a promising approach to overcome the spectrum limitations. In this work we proposed a Deep Learning (DL) approach to learn the channel occupancy model and predict its availability in the next time slots. Our results show that the proposed DL approach outperforms existing works by 5%. We also show that our proposed DL approach predicts the availability of channels accurately for more than one time slot.

This paper presents a Convolutional Neural Network (CNN) approach for classification of low power wide area network (LPWAN) technologies such as Sigfox, LoRA and IEEE 802.15.4g. Since the technologies operate in unlicensed sub-GHz bands, their transmissions can interfere with each other and significantly degrade their performance. This situation further intensifies when the network density increases which will be the case of future LPWANs. In this regard, it becomes essential to classify coexisting technologies so that the impact of interference can be minimized by making optimal spectrum decisions. State-of-the-art technology classification approaches use signal processing approaches for solving the task. However, such techniques are not scalable and require domain-expertise knowledge for developing new rules for each new technology. On the contrary, we present a CNN approach for classification which requires limited domain-expertise knowledge, and it can be scalable to any number of wireless technologies. We present and compare two CNN based classifiers named CNN based on in-phase and quadrature (IQ) and CNN based on Fast Fourier Transform (FFT). The results illustrate that CNN based on IQ achieves classification accuracy close to 97% similar to CNN based on FFT and thus, avoiding the need for performing FFT.

We consider the problem of complementing the capacity of an existing network of macro base stations by dynamically placing a network of 5G small base stations in the form of Unnamed Aerial Vehicles UAV (better known as drones). Our goal is to maximize the capacity boost provided by the UAVs in each considered time frame and extend the battery life of the served mobile users. With this in mind, we propose two clustering algorithms that build on mobile users' spatio-temporal data excess demand (here intended as the portion of demand which is not satisfactory addressed by the existing macro base stations). For the numerical analysis, we use real Beijing downtown trajectory data. The obtained results show that our algorithms perform well and can be considered for enabling real time connection provisioning.

Decentralized cognitive radio networks (CRN) require efficient channel access protocols to enable cognitive secondary users (SUs) to access the primary channels in an opportunistic way Without any coordination. In this paper, we develop a distributed spectrum access protocol for the case where the SUs aim to maximize the total system throughput while competing for spectrum resources. To model the competition amongst SUs, we formulate the spectrum access problem as a {\it distributed welfare game}, in which at each iteration each SU has to compute its marginal contribution to the system's welfare. Moreover, the SUs also need to decide which resource (channel) they should access at the next iteration. To address these challenges, we propose a stochastic learning algorithm based on payoff-based log-linear learning and prove its convergence towards a Pareto-efficient Nash Equilibrium state.

We study the problem of how autonomous cognitive nodes (CNs) can arrive at an efficient and fair opportunistic channel access policy in scenarios where channels may be non-homogeneous in terms of primary user (PU) occupancy. In our model, a CN that is able to adapt to the environment is limited in two ways. First, CNs have imperfect observations (such as due to sensing and channel errors) of their environment. Second, CNs have imperfect memory due to limitations in computational capabilities. For efficient opportunistic channel access, we propose a simple adaptive win-shift lose-randomize (WSLR) strategy that can be executed by a twostate machine (automaton). Using the framework of repeated games (with imperfect observations and limited memory), we show that the proposed strategy enables the CNs (without any explicit coordination) to reach an outcome that: 1) maximizes the total network payoff and also ensures fairness among the CNs; 2) reduces the likelihood of collisions among CNs; and 3) requires a small number of sensing steps (attempts) to find a channel free of PU activity. We compare the performance of the proposed autonomous strategy with a centralized strategy and also test it with real spectrum data collected at RWTH Aachen.

In a cooperative content distribution (CCD) using multiple interfaces, a smart wireless device receives content from a base station (BS) on its cellular interface, and it broadcasts the same content through another wireless interface, such as WiFi. However, different users can experience different link qualities, and users with slow wireless links can be a bottleneck in terms of CCD performance. To address this problem, we propose a device selection method, which leverages multiple interfaces of the selected devices to perform CCD. Our proposed method takes into account the link quality of both primary (cellular) and secondary (WiFi/short-range) interfaces of the devices, and selects the devices with the best link quality for CCD. To analyze the stability of the proposed CCD method against selfish deviators, we model the problem as a repeated CCD game. We show that although the proposed method yields significant gains in terms of energy and frequency carrier savings, it is vulnerable to selfish deviating users. To address this challenge, we propose a carrier aggregation-based incentive mechanism. The analytical and simulation results show that the proposed mechanism maximizes individual and network payoffs, and is an equilibrium against unilateral selfish deviations.

Lattice reduction (LR) is a preprocessing technique for multiple-input multiple-output (MIMO) symbol detection to achieve better bit error-rate (BER) performance. In this paper, we propose a customized homogeneous multiprocessor for LR. Each individual core is based on transport triggered architecture (TTA). We propose a few modifications of the popular LR algorithm, Lenstra-Lenstra-Lovász (LLL) for high throughput. High level programming is used to implement the control path of the TTA cores and several special function units are designed to accelerate the program. The multiprocessor takes 187 cycles to reduce a single matrix for LR. The architecture is synthesized on 90 nm technology and takes 405 kgates at 210 MHz.

Thanks to the smart device revolution, modern wireless devices have increased computational/storage capabilities and can also support for multiple network interfaces such as cellular and WiFi interfaces. Intelligent utilization of multiple network interfaces can address the problem of cellular traffic congestion and it can also increase the frequency resources of cellular networks. Cooperative content distribution (CCD) is one such technique that can be performed by using multiple wireless interfaces. In CCD, a device receives content from a base station on its cellular interface and distributes it to other devices in its vicinity through another wireless interface such as WiFi. However, due to the broadcast nature of the secondary links such as WiFi, even a single bad link can serve as a bottleneck in terms of the CCD performance. To address this problem, in this paper, we propose a device selection method for CCD that takes into account both the primary (cellular) and secondary link (WiFi/short-range) network interfaces. The proposed method incurs little overhead as it utilizes information such as acknowledgement of data packets, that already exists in the network. We evaluate and compare (with the other content delivery methods) the performance of the proposed method in terms of: number of carriers utilized by a cellular base station (BS); average bits-per-Joule performance; and the average time required to deliver a content file. Moreover, we also take into account the impact of the presence of independent competing/interfering links (such as competing users in the unlicensed band) on the performance of the proposed method.

Under the opportunistic spectrum access paradigm, the shared pool of spectrum bands that the multiple autonomous cognitive radios (CRs) need to compete for is not necessarily homogeneous. The non-homogeneity in channels may lead to payoff distribution conflict among autonomous CRs, as each CR would prefer the outcome in which it selects the more desirable channels. To address this challenge, we have designed an adaptive strategy that (without explicit coordination) enables the CRs to autonomously reach an outcome that maximizes the total CR network throughput and minimizes the payoff distribution conflict among the CRs. We utilize the framework of repeated games with private monitoring to: 1) study the dynamic channel selection problem; 2) analyze the stability of the proposed strategy; and 3) investigate the impact of deviations by a selfish CR on the performance of the proposed strategy. In our model, multiple autonomous CRs are not able to observe the channel selections of other competing CRs. Rather, they get a signal from which the selections must be inferred.

We design an efficient sensing order selection strategy for a distributed cognitive radio (CR) network, where two or more autonomous CRs sense the channels sequentially (in some sensing order) for spectrum opportunities. We are particularly interested in the case where CRs with false alarms autonomously select the sensing orders in which they visit channels, without coordination from a centralized entity. We propose an adaptive persistent sensing order selection strategy and show that this strategy converges and reduces the likelihood of collisions among the autonomous CRs as compared to a random selection of sensing orders. We also show that, when the number of CRs is less than or equal to the number of channels, the proposed strategy enables the CRs to converge to collision-free channel sensing orders. The proposed adaptive persistent strategy also reduces the expected time of arrival at collision-free sensing orders as compared to the randomize after every collision strategy, in which a CR, upon colliding, randomly selects a new sensing order.

Although establishing cooperation in a wireless network is a dynamic process, most game theoretic coalition formation models proposed in the literature are static. We analyze a dynamic coalition formation game based on a Markovian model for the spectrum sharing problem in an interference channel. Our model is dynamic in the sense that distributed transmitter/receiver pairs, with partial channel knowledge, reach stable coalition structures (CSs) through a time-evolving sequence of steps. Depending on an interference environment, we show that the game process either converges to the absorbing state of the grand coalition or to the absorbing state of internal and external stability. We also show that, due to myopic links, it is possible that the core of the game is nonempty, but links cannot form the grand coalition to utilize the core rate allocations. We then formulate a condition for the formation of the stable grand coalition. Using simulation we show that coalition formation yields significant gains in terms of average rates per link for different network sizes. We also show average maximum coalition sizes for different distances between the transmitters and their own receivers. Finally, we analyze the mean and variance of the time for the game to reach the stable coalition structures.

In cooperative spectrum sensing, information from several cognitive radios (CRs) is used for detecting the primary user. To reduce sensing overhead and total energy consumption, it is recommended to cooperate only with the CRs that have the best detection performance. However, the problem is that it is not knowna prioriwhich of the CRs have the best detection performance. In this letter, we are proposing three methods for selecting the CRs with the best detection performance based only on hard (binary) local decisions from the CRs. Simulations are used to evaluate and compare the methods. The results indicate that the proposed CR selection methods are able to offer significant gains in terms of system performance.