http://repository.iiitd.edu.in/xmlui/handle/123456789/799 Sat, 11 Apr 2026 13:06:25 GMT 2026-04-11T13:06:25Z http://repository.iiitd.edu.in/xmlui/handle/123456789/844 Coexistence between heterogeneous wireless networks Gopal, Sneihil; Kaul, Sanjit Krishnan (Advisor) The proliferation of data-intensive applications has led to an exponential growth in wireless data traffic over the last decade. Solutions to accommodate this increase in traffic demand include (a) improvements in technology, for example, using high order modulation and coding schemes and multiple-input multiple-output (MIMO) systems, (b) network densification via deployment of small cell networks, and (c) efficient use of spectrum including spectrum refarming and spectrum sharing amongst different networks. While each solution has its own advantages and associated challenges, in this thesis we focus on spectrum sharing between networks. Sharing spectrum bands, which are underutilized temporally or spatially, is a promising strategy to address the demand for spectrum. However, it introduces novel challenges that result from the networks having to coexist with each other. Coexistence could be challenging for several reasons, including disparity in spectrum access rights assigned to the networks by regulatory bodies and differences in technologies and utilities of the networks sharing the spectrum. For instance, in a paradigm shift in the US and Europe, the spectrum licensed to TV operators for exclusive use was opened for use by low power unlicensed devices, provided they did not impair the reception of TV broadcast at TV receivers. More recently, spectrum that was allocated for use by Intelligent Transportation Systems (ITS) was opened up by the Federal Communications Commission (FCC) for high throughput WiFi networks. This resulted in a coexistence scenario where while the networks have equal rights to the spectrum, they care for the different utilities of information timeliness and throughput, respectively. In this thesis, we address in detail the above two scenarios of coexistence for a CSMA/CA based access of the shared spectrum. Motivated by the distinct behavior of the network that cares for information timeliness and the growing interest in real-time monitoring applications, we conclude the thesis with novel insights on spectrum sharing amongst selfish nodes that care for timely delivery of information updates. TV Whitespaces (TVWS) refers to the spectrum licensed for TV broadcast that was opened up by regulators for use by secondary (unlicensed) devices. We investigate the deployment of White-Fi networks of secondary devices, which coexist with TV networks, and the resulting throughputs. White-Fi networks use WiFi-like physical layer and medium access control (MAC) mechanisms. Unlike WiFi networks that operate in the 2.4 and 5 GHz bands and typically have a coverage of up to 100 m, outdoor White-Fi cells have much larger coverage of up to 5 km. As a result, nodes in a White-Fi cell see significant spatial heterogeneity in channel availability and link quality. We model the MAC throughput of a multi-cell city-wide White-Fi network. We formulate a throughput maximization problem for the White-Fi network under the constraint that its nodes’ maximum aggregate interference at TV receivers is within acceptable limits. We propose a heuristic method and illustrate its efficacy over hypothetical deployments of White-Fi networks coexisting with real TV networks in the US cities of Columbus and Denver, which are good examples of heterogeneity in channel availability and link quality in TVWS. Our proposed framework provides useful insights. For instance, we show that while Columbus has higher channel availability than Denver, surprisingly, its network throughput is lower, indicating that more channels may not result in increased throughput. Next, we investigate the coexistence of two networks, one of which cares for information timeliness and the other for throughput. This is motivated by a recent ruling in which the FCC opened up the 5.85−5.925 GHz ITS band, used for vehicular networking, for the unlicensed 802.11ac/802.11ax devices. While both networks have similar spectrum access rights, the incumbents of the ITS band, i.e., vehicular nodes, value timely delivery of information updates, and the sharers, i.e., the WiFi devices, desire high throughput. This novel spectrum sharing scenario raises an interesting question of whether such networks would cooperate or compete for spectrum access. We address this question using a game theoretic approach. We capture the timeliness of information using the metric of age of information. We refer to the network that cares for timeliness as an age optimizing network (AON) and the other as a throughput optimizing network (TON). We study their coexistence under the assumption that both networks share the spectrum using a CSMA/CA based access mechanism and that the AON aims to minimize the age of updates while the TON seeks to maximize throughput. We employ a repeated game-theoretic approach that allows us to answer whether a simple coexistence etiquette that enables cooperation between networks is self-enforceable. Specifically, we introduce a coordination device, which is a randomized signaling device that allows the AON and the TON to access the spectrum in a non-interfering manner. The networks employ a grim trigger strategy when cooperating which ensures that networks would disobey the device only if competition were more beneficial than cooperation in the long run. We apply the proposed etiquette to two distinct practical medium access settings: (a) when collision slots (more than one node accesses the spectrum leading to all transmissions received in error) are at least as large as successful transmission (interference-free) slots, and (b) collision slots are smaller than successful transmission slots. To exemplify, the former holds when networks use the basic access mechanism defined for the 802.11 MAC and the latter is true for networks employing the RTS/CTS based access mechanism. We show that for both medium access settings, while cooperation is self-enforceable when networks have a small number of nodes, networks prefer competition when they grow in size. Our study of coexisting age and throughput optimizing networks shows that an age optimizing network behaves differently from a throughput optimizing one. This motivated us to consider the coexistence of nodes that care for timeliness of information and share the same spectrum. As before, we employ a game theoretic approach. We formulate a non-cooperative one-shot game with nodes as players and age of information as their utilities. We investigate nodes’ equilibrium strategies in a CSMA/CA slot for the aforementioned medium access settings, i.e., when collisions are longer than successful transmissions and when collisions are shorter. For each setting, we provide insights into how competing nodes that value timeliness share the spectrum. We find that access settings exert strong incentive effects. Specifically, we show that under decentralized decision making by nodes, when collisions are shorter, transmit is a weakly dominant strategy, and when collisions are longer, no dominant strategy exists. For the latter case, we analytically derive the mixed strategy Nash equilibrium for when the ages at the beginning of the slot satisfy certain conditions. Tue, 01 Dec 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/844 2020-12-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/840 Highly flexible dual and tri band impedance transformation techniques and their applications in advanced passive and active circuits Banerjee, Deepayan; Hashmi, Mohammad S. (Advisor) The proliferation of multi-band and multi-standard wireless systems has led to frequent advancements in their design techniques and schemes. In essence, the multi-band devices ensure a reduction in the resources required for the development of a multifrequency transceiver, considering that individual components can be utilized to achieve the desired functionalities at multiple frequencies. Furthermore, the operation of one device at different frequencies of interest increases the re-usability of the same. For example, for a tri-band power divider, a single device can split power at three different frequencies of interest, without other additional supporting circuitry. However, the design of multi-band circuits has been found to be extremely challenging. The key constraints are the achievable frequency ratios (or band separation) and the achievable impedance transformation ratios (or impedance gradient). The designs reported in the literature are limited in frequency ratio, which limits their usefulness in applications having widely separated frequency bands. For example, the design of a device to operate at GSM downlink (900MHz) and WiFi-LTE (5.8GHz) is not practically realizable using current techniques due to wide band separation. Another vital factor for multi-band devices is the requirement of “per-band bandwidth” according to the wireless standards. It has been accepted that for closely separated bands, the perband bandwidth is either too low or too-high (essentially making it wideband), and this is another pressing issue requiring attention. The other key challenge of impedance transformation ratio or “Impedance Gradient”, when appropriately addressed, can be potentially useful for numerous applications, including RF Energy Harvesting, Buttler Matrix in Beamforming, etc. Once again, the existing designs and schemes are pretty good for lower transformation ratios but are extremely limited for a wide range of impedance variations. The literature is replete with dual-band architectures, but the reports on tri-/quadand other higher-bands are still in infancy. There have been proposals of some generalized designs, but these have definite limitations. The major limitation of such designs is extremely tedious design procedures and very complicated mathematical formulations when extended to tri-band and above applications. Moreover, most of these generalized techniques rely on convergence technique or graphical approach requiring optimization, thereby essentially leading to a hit-and-trial approach. This issue can be attributed to the absence of closed-form design equations. In addition, most of the existing design techniques are limited in frequency and impedance transformation ratio. The doctoral research, therefore, envisages addressing some of the most important concerns mentioned above. This necessitates the determination of closed-form design equations for proposed circuits, development of simplified design procedures, and miniaturization of the proposed designs. All of these developments include various aspects such as enhancement in frequency and impedance transformation ratios, increased bandwidth per frequency band, and inherent DC blocking ability at all the selected frequencies. In the context of active circuits, the additional directions that are envisaged are the design, optimization, and linearization. For example, multi-band RF Power Amplifiers in IoT applications within an indoor environment need to operate at low power, but it may often require optimization and linearization techniques to achieve decent performance (PAE, IIP3, etc.) at all the frequencies simultaneously. The application focus of this research work is commercial communication, satellite and military, and the upcoming IoT. In all of these applications, the RF front-end modules are one of the key players, and therefore the planned objective of this research work has the potential to significantly advance the current state-of-the-art that may eventually lead to a paradigm shift in the design of such applications. Sun, 20 Dec 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/840 2020-12-20T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/833 Leveraging EMI signals for appliance detection and energy harvesting Gulati, Manoj; Ram, Shobha Sundar (Advisor); Singh, Amarjeet (Advisor) Electromagnetic interference (also known as EMI) is a byproduct of high-speed switching circuits used inside most of present-day electrical and electronic appliances. EMI propagates through conduction along the power lines and through radiation to limited distances. Due to its intrusive nature, EMI signals are generally suppressed or filtered out. Despite this, these signals are fairly ubiquitous. Hence, we explore the possibility of leveraging the weak EMI signals for two applications - appliance detection and energy harvesting. There has been increased research, in recent years, in appliance detection for nonintrusive load monitoring(NILM). NILM facilitates consumers with direct energy feedback, information regarding daily activities, and supports data-driven load scheduling for realizing the long-term goal of optimization of energy consumption in buildings. Traditionally, appliance detection has relied on low-frequency smart meter data. How-ever, in current literature, NILM has been unsuccessful in identifying many information technology loads - such as laptops, desktop computers, modems, and projectors- due to their complex time-varying power consumption patterns. In our thesis, we have investigated the use of conducted and radiated EMI, arising from the switching circuits within these loads, as unique and time-invariant features for detection and classification. Differential mode (DM) conducted EMI signals were first proposed in 2010 as possible features for identifying appliances having complex power consumption patterns. However, these EMI signals were not robustly characterized to ascertain their effectiveness in real world scenarios. In my thesis, we conducted an in-depth study of DMEMI signals through both measurements and simulations for 24 different appliances. Our studies showed that the performance of DM EMI is impacted significantly by the power line impedance, the filters present in the switching power supply circuitry of neighboring appliances on a common power line and power line harmonics. Based on our findings with DM EMI, our follow-up work proposed common-mode(CM) conducted EMI for appliance detection. CM EMI originates from capacitive coupling from the switching circuitry and flows along the earth conductor. Hence, the signal is not affected by power line harmonics. Also, most appliances are not fitted with common mode chokes because of which the signals from multiple appliances do not interfere with each other. Hence, the CM EMI is a far more robust feature for appliance detection. In order to experimentally test our hypothesis, we designed an EMI sensor to simultaneously monitor both DM and CM EMI from appliances. We evaluated the detection performance, across multiple instances of five commonly used electronic appliances typically found in office setups. We used statistical features de-rived from the histograms of measured EMI signals to differentiate across the various classes of appliances. We found that the CM EMI indeed serves as a superior feature, having higher detection accuracy of 87% in comparison to lower accuracy of 45% in the case of DM EMI. Expanding on this line of work, we envision CM EMI data to be combined with instantaneous, low-frequency power data gathered from smart meters to provide actionable insights to energy stakeholders. Along with conducted EMI, we also explored radiated emissions (also known as RFI) from appliances with an end goal of providing personalized energy apportionment(PEA). PEA is a process of attributing energy consumption to individual stakeholders in a shared space. As radiated emissions can propagate as far as 30cm, they can be leveraged using a wearable sensing device for mapping appliance usage to the instantaneous power data from smart meters. In our study, we characterized RFI from 10 electrical and electronic appliances, in multiple test scenarios, at variable distances. Our test setup consisted of custom-off-the-shelf components like software defined radio and ultra-wideband antennas. We found that a simple peak finder algorithm yielded 72% accuracy for detecting these appliances using RFI signals. Taking our initial exploration with EMI signals one step further, we employed low-frequency stray emissions from AC power lines for energy harvesting. Energy harvesting is a process of scavenging energy from ambient physical sources - such as mechanical load, vibrations, temperature gradients, and light - to support battery-less low-power sensing in the nW-mW range. Since the advent of cyber-physical systems and the internet of things, energy harvesting has been a topic of interest. However, the intermittent nature of existing natural sources restricted the applications of energy harvesting. In this thesis, we leveraged the ubiquitous and continuous nature of stray electric fields from power lines, for facilitating 24x7 energy harvesting for long-term, self-powered deploy and forget sensor networks. Stray electric field signals do not require an isolating wire bundle or an active appliance for harvesting, unlike stray magnetic field signals. We proposed a novel capacitive energy harvester (CapHarvest) with an ultra-low-power management circuit connected to the harvesting electrodes to effectively gather energy from this nano-watt source. Furthermore, we demonstrated the efficiency of our circuit for powering two applications. The first application, called Appliance Tag, is a new stick-on sensing system which monitors appliance state using stray magnetic field signals present around the power line. The second application, called Farm Check, monitors all the ambient physical parameters like temperature, light intensity, and relative humidity for vertical farming applications. This thesis paves a new dimension of sensing and repurposing the otherwise ignored ubiquitous EMI signals for appliance detection and energy harvesting to support the long-term goal of energy sustainability. In the future, the blend of simultaneous sensing and energy harvesting - as demonstrated with Cap Harvest - may enable more such exciting applications. Wed, 01 Jan 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/833 2020-01-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/832 Exploring geometric constraints for learning representations for visual data Shukla, Ankita; Anand, Saket (Advisor) Representation of visual data is a connecting link between the perceptual world and machine based processing. Over the decades, the computer vision community is dedicated to improving these representations, so that it can assist humans in a wide range of applications from medical imaging to visual search and face recognition systems to name a few. In this thesis, we explore geometric constraints to aid in learning representations for various computer vision applications that either have access to only limited amount of labeled training data, abundant unlabeled training data or a combination of two. We investigate two types of geometric constraints: manifold and semantic. The contribution in this thesis can be categorized into two parts based on the geometric constraints used. In the first category, we use the geometry of manifolds. First, we use the geometry of Stiefel manifold to learn a linear transformation of feature representations in the supervised setting, and we show improved generalization in low training-data settings. We also show the same manifold constraint to be effective in the unsupervised learning of disentangled representations, which can help improve the interpretability of deep networks. The third problem is that of defense against adversarial attacks on deep networks. Using the geometry of the Grassmann manifold, we show that our subspace based representations of an adversarially perturbed input sample are sufficiently close to their clean counterparts, and can serve as adefense strategy without the need of any retraining or fine-tuning of the network. In the second category, we make use of semantic constraints and derive a loss term that leverages the statistical manifold, i.e., the space of probability distributions. We apply this loss term in two learning scenarios. First, we use it to combat over-fitting in supervised representation learning in case of limited labeled training data for visual animal biometrics task. We show that it improves the robustness and generalization of the representations for primate face recognition as well tiger re-identification problem. Secondly, we use it for learning cluster able representations in a semi-supervised setting, where it has access to limited labeled data along with abundant unlabeled data. In this thesis, based on the improvements across different applications and settings, we conclude that the geometric information is useful for visual data representation learning regardless of the level of supervision. Thu, 01 Oct 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/832 2020-10-01T00:00:00Z