Edge Analytics and Intelligent Systems
Relevant Publication: COMSNETS '17, SECON '16, TWC '15, INFOCOM '15

Crowdsourced enforcement of spectrum policies

The inherent unguided and unbounded nature of radio wave propagation, allows spectrum infractions to cause widespread damage and makes it hard to locate at the same time. In this work [TWC '15], we leverage omnipresent mobile users (edge devices) to implement a paradigm of eye-witness for detecting violations of spectrum policies. The key challenge is to find the trustworthy and accurate field measurements. A Complementary problem is planning of autonomous agents for signal detection and localization, which is investigated in this latest work. A combination of greedy, path prunning and shortest path routing algorithm is shown to achieve 3-approximation ratio in real deployments.

Predictive analytics for non-stationary V2X channels

Vehicle to Road-side infrastructure wireless channels are predictable because the vehicle path, the large scale scattering, fading and shadowing remain static with time. However, small scale scattering from neighboring vehicles (density dependent) and Doppler shift leads to variation of the channel response for a stretch of road. In this work [COMSNETS '17], we employ a prediction based on topological data provides a better view of the channel and the Access Points (APs) are more suited for this type of analytics.

Coexistence of heterogeneous networks

This paper [SECON '16] examines the important question of how should a user optimally select the best access network in a HetNet with mmWave to maximize throughput and minimize switching costs? Modeling mmWave radio access technologies (RATs) as a stochastic 3-state process based on their physical layer characteristics, we develop a UE-based online learning-based approach to balance between exploration and exploitation. Using a Multi-Arm Bandit Problem formulation, the algorithm minimizes the total regret associated with network swithces and achieves 24% increase in total throughput compared to existing techniques for high throughput mmWave RATs that vary over a fast timescale.


Spectrum aware video streaming for Whitespaces

The recent three tier shared licensing model proposed by the Federal Communications Commission (FCC) defines users with different access rights to the same band of frequencies leading to more dynamic form of adaptation by the user, specifically the Tier-3. In this work [INFOCOM '15], we propose a Markov Decision Process that guides the user to stream video over Whitespaces. With the knowledge of the arrival processes of higher priority users, along with real-time channel quality measurements, a Tier-3 user chooses the optimal bitrate for a video segment.

Harnessing Software Defined Radios for Networking
Relevant Publication: SIGCOMM '09, INFOCOM '09, TMC '13, CROWNCOM '12, IHMMSEC '12, DySPAN '10

Low latency physical layer signaling

In SMart ACKnowledgment (SMACK) [SIGCOMM '09], we have successfully demonstrated the use of OFDM subcarriers to perform low latency signaling in wireless networks. With an appropriate application interface, simultaneous transmission can greatly speed up common group communication primitives. We demonstrate how simultaneous transmission can be used to implement a reliable broadcast for an infrastructure network in an SDR testbed. The PHY Aided MAC (PAMAC) [INFOCOM '09] protocol is an example of utilizing OFDM subcarriers to have clients signal whether they have packets to send and their queue information. Again, using the programmable physical layer, the AP schedules the clients efficiently, while wasting little of the spectrum on signaling overhead.

Multiuser communication using OFDM

In GRaTIS or Group Rate Transmission with Intertwined Symbols [TMC '13], we describe, evaluate and implement a practical multiuser communication scheme that exploits discrete "steps" in modulation rates to transmit two packets in the time normally needed to transmit a single packet, increasing the aggregate throughput . In another project, the flexible PHY is used to implement a high capacity covert channel as seen in steganographic applications [IHMMSEC '12]. By injecting pre-distortion in the OFDM modulation constellation, the signal is made to mimic a noisy version of an ideal signal, which is typically attributed to the channel noise or certain hardware impairments. We implemented this on the prototype hardware and validated the steganographic strength using various radio measurements and statistical signal analysis.

Dynamic spectrum access and NC-OFDM

The SDR prototype has been used to solve practical synchronization problem for non-contiguous OFDM [DySPAN '10]. Unlike the contiguous channelized access model, wideband non-contiguous access posses a critical challenge of synchronization. In this work, we propose a practical algorithm and hardware implementation to overcome this challenge. Equipped with this blind synchronizer, we propose a MAC layer design to enable flexible channel access while achieving co-existence with the incumbent and other secondary heterogeneous networks.



SDR Prototype and Architectures for Cognitive Radios
Relevant Publication: WICON '08, ANCS '10, SRIF '13

Prototype hardware for wireless research

In order to make SDRs useful for networking research, it has to have the appropriate architecture. The first step towards making SDR relevant to crosslayer networking research is to build a 802.11a/g protocol compliant, OFDM physical layer prototype that is fully compatible with commodity WiFi devices. The contribution of this work is in synthesizing the signal processing algorithms used in OFDM based PHY while making them programmable [WICON '08]. With various concurrent wireless protocols in mind, we define a set of fundamental operations, which a generic OFDM transceiver should support, in order to operate as a Software Define Cognitive Radio (SDCR) [ANCS '10]. Re-defining the radio physical layer, beyond a fixed-function operation is motivated by MAC-PHY crosslayer research and cognitive networks.

Composability of SDR

Composing On-Demand Intelligent Physical Layers or CODIPHY [SRIF '13] solves two fundamental problems in practical cognitive radio networks: collaboration between two radio PHY layers with varying capabilities to agree on a common communication protocol, using an ontology based description of the internal structure of the radio subsystems and secondly, provide a method to compose a functioning radio pipeline from a set of pre-compiled components, using the high-level dataflow represented by the ontology, to target heterogeneous platforms. Using suitable back-end compilation, the ontology is translated to the preferred implementation language required by the target platform.