-
*
City Infrastructure Assessment
(1) Anomaly Detection for IoT
This research is studying various approaches to anomaly detection in networks of sensors, such as Internet of Things (IoT) nodes. The goal is to improve on the state of the art in AD using various machine learning and deep learning approaches to Anomaly Detection. Approaches studied previously are simple linear classifiers (such as support vector machines), and clustering approaches (such as K-Means and DBSCAN). Current research is focusing on prediction-based approaches and generative AI to perform AD. Work utilizes the robust Chicago Array of Things dataset; an unlabeled dataset composed of hundreds of sensors spanning a 2-year timeframe. This is Kyle's main interest.
Code Papers BibTeX(2) Sidewalk Assessment and Routing
The scope of the project is to discover a fully automatic approach to assess the underlying features of the sidewalk. Today, sidewalk inspection is done through a series of manual processes, which rely on the subjectiveness of the inspectors and slow maintenance. This project focuses on 1) detecting sidewalk features and distress 2) analyzing the shape of detected features and distress 3) verifying the ADA-PROWAG compliance level of the features and distress. This project is in the process of creating an aerial-ground-multisensor fusion sidewalk dataset. This is Chan's main interest.
Papers BibTeX -
*
Smart and Personalized Routing
Supports the evaluation of personalized routing queries with the following merits. (1) PreGo finds the best route according to a user’s set of personal preferences, e.g., time, distance, safety, attractions, etc. (2) PreGo does not depend on a static snapshot of the underlying road network. Rather, it computes the best route taking into consideration various cost elements at different time instances of the day. (3) For not in a rush user, PreGo recommends a trip start time such that user’s preferences are best fulfilled. (4) Personalization parameters are controlled by each user’s preference weights.
Code Papers BibTeX -
*
Spatial Predictive Queries
A fundamental category of location based services relies on predictive queries which consider the anticipated future locations of users. Predictive queries attracted the researchers’ attention as they are widely used in several applications including traffic management, routing, location-based advertising, and ride sharing.
(1) Predictive Tree (P-tree)
An index structure developed to find the objects predicted to be around a given node in the underlying road network graph after some future time...
(2) Panda
Panda is a framework that supports predictive spatial queries against moving objects on Euclidean space.
(3) ? (In progress)
Code Papers BibTeX
My research interests are in big data, spatial databases, and AI, with applications in smart city and smart health.
-
*
Medical Data Analysis
Develop Machine Learning techniques to discover patterns of interest from health/disease related data.
(1) Reovirus
This work focuses on classifying variants of turkey reoviruses by leveraging various machine learning algorithms such as Support Vector Machines, Naive Bayes, Random Forest, Decision Tree, and deep learning algorithms including Convolutional Neural Networks. This is Maryam's main interest.
Code Papers BibTeX(2) Alzheimer Disease Identification
Explore the capabilities of a diverse set of machine learning techniques to predict the progression of AD. Optimized models are extended to provide different XAI features which support the interpretation of the decisions predicted by the machine learning algorithms. This is Chamudi's main interest.
Coming Soon -
*
Internet of Moving Things
RxSpatial Library
is developed in collaboration with researchers from Microsoft Seattle to provide real-time processing for spatial operations on moving objects. The aim is to support the novel spatio-temporal applications in the era of IoT. Code Papers BibTeX -
*
Management and Analysis For
Collaborating with researchers and scientists from the University of Virginia and University of Pennsylvania in several research threads under the umbrella of smart cities.
(1) Dynamic Taxi dispatch
We design an efficient algorithm for constructing distributional uncertainty sets of random demand vectors, and leverage a quad-tree dynamic region partition method for better capturing the dynamic spatial-temporal properties of the uncertain demand. The average total taxis idle distance is reduced by 20%. This is about 60 million miles or 8 million dollars cost reduction annually for all taxis in NYC compared to solutions based on static grid partition.
(2) Safety and Security
In this work, our goal is to add a layer on the exising smart cities infrastucture to ensure safety and security of the city main services and sensors. The work is still in its early phase.
Code Papers BibTeX -
*
Road Network Compression
COMA
Motivated by the new era of tiny smart devices, (e.g., smart watches), we developed the COMA system for compressing road networks while preserving the quality of the compressed map to fits well with the common spatial operations. COMA algorithm is O(|N|log|N|) and can achieve about a 75% compression-ratio while preserving high quality of the compact map.
Papers BibTeX