Practical Design and Orchestration of Frequency-Shifting RIS for NLoS mmWave Sensing

Published in ACM MobiHoc 2025

mmWave radar powers next-generation sensing in ISAC systems but struggles in non-line-of-sight environments. PRISM breaks through these limits, using RIS-based spatial modulation to turn angular cues into frequency shifts—boosting resolution without heavy computation or tight synchronization. The result: faster, more adaptable radar sensing that works where others fail.

Download here

Scalable Network Tomography for Dynamic Spectrum Access

Published in IEEE INFOCOM 2024

NeTo-X is a scalable network tomography framework that maps interference patterns and estimates joint client access statistics with only linear overhead—paving the way for more efficient dynamic spectrum access in future networks. By exploiting multi-channel diversity and the spatial locality of interference, NeTo-X accurately infers interference statistics from simple pairwise client measurements. Its effectiveness is demonstrated in resource management and jammer localization, where it outperforms baselines and approaches optimal performance while remaining highly scalable.

Download here

Renting Edge Computing Resources for Service Hosting

Published in EAI VALUETOOLS 2023

This paper is about Better-Late-Than-Never (BLTN), an online algorithm which decides the state of an edge computing server based on historic request arrival rates. In this work, we propose a deterministic online policy and characterize its performance for adversarial and stochastic request arrival processes. We also characterize a fundamental bound on the performance of an arbitrary deterministic online policy. Further, we compare the performance of our policy with suitably modified versions of existing policies to conclude that our policy is robust to temporal changes in the intensity of request arrivals.

Download here

Learning Unsupervised Representations for Sensing Humans with mmWave Radars

mmWave radar data is rich in information related to humans like pose and movement. We also observe that collection of data is easy, however labelling the data is extremely hard. In this project, we have built an unsupervised framework to extract spatial and temporal features of the data which are critical to downstream supervised tasks like pose estimation, tracking and identifying individuals. Using the latent space features, we are able to build more effective supervised learning frameworks that require 50 times fewer training samples than prior work.

Backscatter for Low-power IoT Environmental Sensing

In this project, we built an ultra-low power backscatter device for monitoring Atlanta’s urban heat islands, using digital communications techniques to extend the range between the tag and transmitter/receiver by sacrificing throughput. We explore digital range-increasing techniques such as forward error correction (FEC) and investigate low-power digitization and signal generation. We build a system consisting of a co-located transmitter and receiver capable of reading temperature data from a single backscatter tag.

Topological Methods for Data-Driven Analysis

Applied persistent homology, a toolset to analyse high-dimensional data using topological data analysis, to motor control, which includes stride-to-stride fluctuations and gait dynamics, and neurodegenerative diseases.