My research interest lies broadly in the area of Wireless Communications and Networks.
Faster, reliable, and resource-efficient transport of information from one point to another has been at the core of most technological advancements in the past few decades. Wireless communication has thus paved the way for ubiquitous connectivity in modern society. With 5G networks already deployed in several parts of the world, wireless research endeavors are now pushing the boundaries of what 6G and beyond systems would look like in terms of their capabilities, functionalities, and challenges. My vision is to enable the next generation of wireless systems & networks and redefine what a communication network could achieve. The crucial question to address is - what more can a cellular base station or a Wi-Fi Access Point do besides just reliable transport of bits? A search for an answer to this question has led to several research avenues pertaining to joint communication and sensing (JCAS) systems which would make future networks multi-function. More so, with the advent of massive MIMO, intelligent metasurfaces, distributed MIMO, FR2 and FR3 spectrum bands, the physical layer characterization of transmit and receive domains is ever-evolving and presents its own set of challenges for the most efficient system design keeping in mind the distinct quality of service (QoS) requirement of different communication and sensing applications. The endless possibilities with such systems, and their associated challenges excite me, and I try to find the solutions to such challenges through my research. I frequently employ tools from Tensor Algebra, Information Theory, Estimation Theory, Distributed Optimization, and Machine Learning in my research, while also conducting experiments to validate the proposed theory.
Click here for my Google Scholar profile to see my latest publications.
Some of my selected research projects are listed below.
Tensor modeling of multi-domain communication systems
Modern communication systems are inherently multi-domain in nature where they employ signals and systems with features spanning multiple domains for information transmission and reception. These domains can include space, time, frequency, users, code sequences, propagation delay, and varied transmission media, to name a few. Designing modulation schemes and signal processing techniques that can simultaneously benefit from the distinct nature of all the domains is a challenging task in emerging communication systems. Thus, we developed an integrated mathematical approach that simultaneously and mutually exploits all the information transmission domains using tensors, as part of my Ph.D. thesis which can be found here.
Shannon capacity of tensor channels
We derived the Shannon capacity and input power allocation for higher-order tensor channels under a family of power and input constellation constraints. Owing to the multi-domain nature of the input signals, the power constraints in multi-domain communication systems can span one or more domains. My research demonstrates the tensor framework's distinct ability to mathematically model a variety of such domain-specific power constraints. The tensor-based formulation encapsulates the impact of various domains and thus allows collaborative multi-domain precoding and power allocation, and brings forth the inherent trade-off between multiple domains. Several examples of multi-antenna, multi-carrier, and multi-user systems are presented to illustrate the tensor handling of multi-domain interactions. Key results can be found in
Pandey D, Leib H. The tensor multi-linear channel and its Shannon capacity. IEEE Access. 2022
Pandey D, Leib H. Capacity Performance of Tensor Multi-Domain Communication Systems With Discrete Signalling Constellations. IEEE Open Journal of the Communications Society. 2023.
Pandey D, Leib H. Shannon capacity of tensor channels under a family of power constraints. 30th Biennial Symposium on Communications (BSC) 2021.
Tensor-based signal processing in multi-domain communication systems.
The proposed tensor model is further exploited to develop a framework for MMSE estimation methods for multi-domain signals and data. Both proper and improper complex tensors are addressed by the framework. The traditional linear and widely linear MMSE estimators are extended to the tensor setting, resulting in multi-linear and widely multi-linear MMSE estimation. Also, we developed a tensor partial response signaling (TPRS) method, which allows the introduction of controlled interference across domains for spectral and cross-spectral shaping. Key results be found in
Pandey D, Leib H. A tensor framework for multi-linear complex MMSE estimation. IEEE Open Journal of Signal Processing. 2021.
Pandey D, Venugopal A, Leib H. Multi-domain communication systems and networks: A tensor-based approach. Network. 2021.
Pandey D, Leib H. Tensor multi-linear MMSE estimation using the Einstein product. In Advances in Information and Communication: Proceedings of the 2021 Future of Information and Communication Conference (FICC), 2021. recipient of the Best Student Paper Award at FICC 2021.
Pandey D, Leib H. A tensor-based precoder and receiver for MIMO GFDM systems. IEEE International Conference on Communications (ICC) 2021.
Tensor algebra for multi-linear system representation
Tensor algebra as a tool is extremely significant for modeling any multi-domain system, not just communication systems. Thus, we developed the multi-linear system theory using tools from tensor algebra, particularly tensor contraction. This allows us to generalize various linear algebra tools such as eigen value and singular value decompositions to a higher-order setting. Further, we also studied the notions of complex random tensors and how the statistical properties span across the modes of tensors. In particular, different correlation structures for higher-order signals are analyzed. Key results are available in
Pandey D, Venugopal A, Leib H. Linear to multi-linear algebra and systems using tensors. Frontiers in Applied Mathematics and Statistics. 2024.
Pandey D, Decurninge A, Leib H. An Introduction to Complex Random Tensors. IEEE Access, 2024.
Exploiting multi-bounce for enhanced radar imaging
Radar Imaging has garnered fresh interest in the research community due to developing automotive applications, and sensing capabilities that can be built on top of the existing communication infrastructure by leveraging communication waveforms for sensing purposes as well. In this regard, multipath reflections, which have traditionally been seen as a nuisance, hold the promise of enhancing the radar's detection and localization performance. In its most basic essence, multipath essentially provides additional information that can be used to mimic a virtual synthetic aperture, thereby expanding the field of view and resolution limits of the radar. A few specific projects that we are currently working on are as follows:
Beyond field-of-view detection using ambient reflections
mmWave radar has a limited imaging field-of-view due to high directionality and reliance on single-bounce scatter from objects being imaged. We proposed exploiting natural multi-bounce scattering in the environment to enable mmWave radar imaging of objects beyond the single-bounce field-of-view (e.g., around corners and behind the radar). Our implementation on a commercial millimeter-wave MIMO radar testbed demonstrates 2×-10× improvement in the median localization error for humans standing outside the radar’s field-of-view in various indoor and outdoor scenarios. The results have been accepted for publication in ACM MobiCom 2024.
Mehrotra N, Pandey D, Prabhakaran A, Liu Y, Kumar S, Sabharwal A, Hydra: Exploiting Multi-Bounce Scattering for Beyond-Field-of-View mmWave Radar. In Proceedings of the 30th Annual International Conference on Mobile Computing and Networking (ACM MobiCom '24).
Velocity estimation using MIMO radar
MIMO radars can estimate radial velocities of moving objects, but not their tangential velocities. We propose a method of exploiting multi-bounce scattering to enable estimating a moving object’s entire velocity vector – both tangential and radial velocities. We tested our proposed approach using simulations and experiments with TI’s mmWave MIMO radar, AWR2243 cascade. Our initial results have been published in IEEE CISA 2024.
Mehrotra N, Pandey D, Madhow U, Mostofi Y, Sabharwal A. Instantaneous Velocity Vector Estimation Using a Single MIMO Radar Via Multi-Bounce Scattering. IEEE Conference on Computational Imaging Using Synthetic Apertures (CISA) 2024.
Mehrotra N, Pandey D, Madhow U, Mostofi Y, Sabharwal A. Single-Frame MIMO Radar Velocity Vector Estimation via Multi-Bounce Scattering", IEEE Transactions on Computational Imaging 2025.
Detection under the rubble using UWB radar for disaster management
In this project, our aim is to employ UWB radar to detect human presence in complex environments where individuals may be trapped under the rubble of a natural disaster site. Our strategy involves translational and rotational motion of the radar to capture more spatial information, effectively creating a multiview synthetic aperture that improves the localization and detection of closely placed targets. We consider a channel model incorporating the multipath effects on human sensing under the debris in our analysis, and extensive experiments are underway.
A. Deshwal, S. Sinha, A. Azizi, D. Pandey, N. Mehrotra, A. Pal, A. Sabharwal, Through-the-Wall Multi-Person Localization using Translation and Rotation Synthetic Aperture Radar", in IEEE ICASSP 2025.
Resource allocation in massive MIMO networks
SLA-aware scheduling in massive MIMO
Radio Access Network (RAN) slicing virtualizes the infrastructure of wireless networks and partitions it into slices (or enterprises), tailored to fulfill specific use-cases. A key focus in this context is the efficient radio resource allocation to meet various enterprises' service-level agreements (SLAs). In this project, we introduce a channel-aware and SLA-aware RAN slicing framework for massive MIMO networks where resource allocation extends to incorporate the spatial dimension available through beamforming. Initial results have been accepted for publication in ACM CoNEXT 2024, and more extensive testing of the proposed algorithm using RENEW platform at Rice is underway.
An Q., Pandey D, Doost-Mohammady R, Sabharwal A., Shakkotai S., Helix: A RAN Slicing Based Scheduling Framework for Massive MIMO Networks. In ACM CoNEXT 2024, to be published in Proceedings of the ACM on Networking (PACMNET)
Learning-based scheduling for non-stationary MIMO channels
Learning-based MIMO scheduling methods have emerged as promising solutions in multi-user MIMO resource allocation. Based on pre-trained ML models, these methods can provide fast scheduling decisions without solving any optimization problem per transmission. Recent works have used Deep Reinforcement Learning-based methods for MIMO scheduling based on information on channel conditions with very promising results. However, such methods are sensitive to changes in the channel environment, as dynamic variations in channel distributions necessitate retraining the machine learning model. Consequently, the frequent need for retraining increases implementation overhead, whenever the channel distribution changes. In this project, we aim to build a learning-based scheduling approach where we can adapt the policy for a new environment based on policies learned for other environments. As the project progresses, we aim to show that a cooperative scheduling scheme across base stations can be designed where one base station can learn from the already learned process of its neighboring base station.
Beamforming in massive MIMO
Distributed multi-user beamforming in the presence of CFO
Implementing distributed multi-user beamforming requires achieving accurate over-the-air timing and frequency synchronization among distributed access points (APs), particularly due to residual frequency offsets caused by local oscillator (LO) drifts. In this project, we do an analytical, simulation, and experimental assessment of distributed beamforming methods in the presence of frequency synchronization errors. We characterize SINR as a function of channel characteristics and statistical properties of carrier frequency offset (CFO) among AP antennas. Through experimental evaluations conducted with the RENEW massive MIMO testbed, we collected comprehensive datasets across various experimental scenarios to assess the performance of conjugate beamforming and zero forcing in the presence of CFO and validate our analysis.
Zafari M, Pandey D, Doost-Mohammady R. Experimental Assessment of Distributed Multi-User Beamforming in the Presence of CFO. draft submitted to IEEE Transactions on Vehicular Technology 2024.
Distributed optimization for beamforming in cell-free massive MIMO systems
In cell-free massive MIMO systems with multiple distributed Access Points (APs) serving multiple users over the same time-frequency resources, downlink beamforming is done through spatial precoding. Precoding vectors can be optimally designed to use the minimum downlink transmit power while satisfying a quality-of-service requirement for each user. In this work, we formulate a multi-user beamforming optimization problem to minimize the total transmit power subject to per-user SINR requirement and propose a distributed optimization algorithm based on the Alternating Direction Method of Multipliers (ADMM) to solve it. The work has been accepted for publication in Asilomar 2024.
Zafari M, Pandey D, Doost-Mohammady R, Uribe C. ADMM for Downlink Beamforming in cell-free massive MIMO systems, in Asilomar Conference on Signals Systems and Computers 2024.
Experimental study of AoA estimation for UAV to massive MIMO.
In this project, we evaluated the performance of Angle of Arrival (AoA) estimation algorithms in Unmanned Aerial Vehicles (UAV) communication networks utilizing massive Multiple-Input Multiple-Output (MIMO) base stations. Five different AoA estimation algorithms were evaluated and the results show the impact of under-sampling on AoA estimation, specifically in the detection of multi-path with higher normalized power. The effects of azimuth AoA estimation via horizontal subarrays and the impact on multi-path AoA estimates for hovering drones were also examined. The work was presented in MILCOM 2023
Rice T, Pandey D, Ramirez D, Knightly E. Experimental Evaluation of AoA Estimation for UAV to Massive MIMO. IEEE Military Communications Conference (MILCOM) 2023
Exploiting degree-of-control in transmissive metasurfaces
In this project, we focus on a pattern-reconfigurable transmissive RIS-assisted multiuser MISO downlink communication system. The reconfigurability of RIS can be exploited to perform analog beamforming. Thus, the digital baseband precoder needs to be simultaneously optimized along with the RIS parameters to achieve the optimal combined beamforming effect. Thus, we study the problem of jointly optimizing the digital precoder and the phase and pattern parameters on the RIS. We combine DRL based approach with conventional methods to design the optimal scheme and analyze the trade-off between phase and pattern parameters under DoC constraints at the RIS. The main question we aim to answer is that given certain constraints on the rate of control, what is the optimal load-balancing between the analog front end and the digital baseband to execute beamforming.