Intelligent Systems
  • Agent-based Scenario Planning for a Smart & Connected Community against Sea Level Rise in Tampa Bay:

    Sea-level rise and flooding pose significant risks to communities and infrastructure in Florida and many other coastal states. The work in this proposal seeks to engage a variety of stakeholders in the Tampa Bay region, from citizens to businesses to government agencies, in exercises grounded in Big Data analytics and agent-based modeling that simulates the dynamics and uncertainty of responses to sea level change. The ultimate objective is to create a more connected community by convening these stakeholders and having them work together towards crafting resilient responses to the possible outcomes of sea level rise scenarios.

    Project Website

    Collaborators: Mark Hafen (USF, Urban Planning), Alan Bush (USF, Honors College), Ryan Carney (USF, Advanced Visualization Center & Integrative Biology)
    Sponsor: NSF

  • Towards Smart & Autonomous Buildings:

    To realize the overarching vision of smart cities, there is a need to transform the today's largely inanimate buildings into smart and autonomous living environments. Such smart and autonomous buildings (S&ABs) are expected to be intelligent physical systems that can continuously adapt their functions to cater to their residents' needs and behaviors. By exploiting such features, S&ABs will be able to enhance energy efficiency, security, privacy, and the overall quality of life in tomorrow's smart cities. However, the practical deployment of such S&ABs is contingent upon meeting several challenges that range from endowing them with artificial intelligence features to enabling autonomous decision making across their various subsystems.

    Collaborators: Ismail Guvenc (NCSU, ECE), Walid Saad (Virginia Tech, ECE), Kemal Akkaya (FIU, ECE), Wangda Zuo (U Colorado Boulder, Civil Engineering), Aslihan Akkaya (FIU, Anthropology)

  • Intelligent Transportation Systems:

    Increasing urbanization worldwide looks likely to be a trend to continue in the foreseeable future. It brings in many complexities in energy consumption, transportation, housing, and, in more general, living experience of the urban dwellers in terms of pollution, waste, noise, congestion, etc. At the same time, these challenges push new opportunities to the forefront in new innovation ecosystems. One such promising framework is the paradigm of Smart Cities, where the power of computational techniques is utilized for more optimized use and management of city’s assets to improve urban living. An important subsystem for the success of Smart Cities is its transportation infrastructure by means of an intelligent transportation system (ITS). The defining characteristic of ITS is its data-driven methodology.

    Collaborators: Suleyman Uludag (U Michigan Flint, Computer Science), Esma Dilek (Istanbul Metropolitan Municipality)
    Student: Ammar Haydari (USF)

  • Autonomous Driving:

    We consider the problem of recognizing events around a (autonomous) vehicle using the vision and mm-Wave radar sensors. Event recognition involves not only recognizing the object, but also the action the object is taking. Examples of an event are a traffic light turning red/green, a pedestrian crossing the road, the road edge curving right/left, a parked car, and a passing by car. We propose to first sense the environment using the tailored capabilities of mm-Wave radar; and then recognize the event by fusing (i.e., jointly analyzing) the data from the vision and mm-Wave radar sensors. For fusion, we follow an event-based methodology. Specifically, we model the process how an unobserved event generates our multimodal observations via a generative latent variable; and statistically infer about the latent variable denoting the event type.

    Collaborators: Kwang-Cheng Chen (USF, EE)

Network Analysis

  • Social Networks:

    Event detection in a multimodal Twitter dataset is considered. We treat the hashtags in the dataset as instances with two modes: text and geolocation features. The text feature consists of a bag-of-words representation. The geolocation feature consists of geotags (i.e., geographical coordinates) of the tweets. Fusing the multimodal data we aim to detect, in terms of topic and geolocation, the interesting events and the associated hashtags. To this end, a generative latent variable model is assumed, and a generalized expectation-maximization (EM) algorithm is derived to learn the model parameters. The proposed method is computationally efficient, and lends itself to big datasets.

    Collaborators: Alfred Hero (U Michigan, ECE)
    Representative papers: "Multimodal Event Detection in Twitter Hashtag Networks", Journal of Signal Processing Systems, 2016

  • Gene Networks:

    We consider two datasets. First one is micro-array data of gene expression levels for a given time frame or different states of the cell. Binary adjacency matrix A accounts for gene co-expressions. Second dataset is a matrix of contacts counts between gene pairs (called Hi-C contact map) where entries are function of the spatial distances. As a result, we have the weighted adjacency matrix H of the spatial network of genes. The overall objective is to infer the different network modules from networks given by A and H assuming they have some common modules, and identify common modules and private modules.

    Collaborators: Alfred Hero (U Michigan, ECE), Sijia Liu (U Michigan, ECE), Indika Rajapakse (U Michigan, Bioinformatics)
    Student: Hafiz Tiomoko Ali (CentraleSupelec, France)

  • Multimodal Data Fusion:

    We are interested in learning, in an unsupervised fashion, a low-dimensional feature vector for the observed objects from high-dimensional heterogeneous data. The learned feature vector can then be used for various machine learning tasks, such as clustering, classification, regression, etc. Consider a heterogenous system of P objects (i.e., instances) each of which generates multimodal data in the form of M disparate data types. Data from each modality (i.e., data type) can be modeled with a probability distribution from the exponential dispersion family (e.g., Gaussian, Poisson, multinomial), which is a generalization of the exponential family. Under a “big data” setup, in which the number of instances P, the dimension of the data vector and possibly the number of modalities M are large, our objective is to learn a feature vector for each object from the multimodal dataset.

    Collaborators: Alfred Hero (U Michigan, ECE)
    Representative papers: "Multimodal Factor Analysis", IEEE International Workshop on Machine Learning for Signal Processing, 2015

  • Anomaly Detection:

    We consider the online and nonparametric detection of abrupt and persistent anomalies, such as a change in the regular system dynamics at a time instance due to an anomalous event (e.g., a failure, a malicious activity). Combining the simplicity of the nonparametric Geometric Entropy Minimization (GEM) method with the timely detection capability of the Cumulative Sum (CUSUM) algorithm we propose a computationally efficient online anomaly detection method that is applicable to high-dimensional datasets, and at the same time achieve a near-optimum average detection delay performance for a given false alarm constraint. We provide new insights to both GEM and CUSUM, including new asymptotic analysis for GEM, which enables soft decisions for outlier detection, and a novel interpretation of CUSUM in terms of the discrepancy theory, which helps us generalize it to the nonparametric GEM statistic.

    Student: Mahsa Mozaffari (USF, EE)
    Sponsor: SCEEE
    Representative papers: "Online Nonparametric Anomaly Detection based on Geometric Entropy Minimization", IEEE International Symposium on Information Theory, 2017

  • Smart Grid Security:

    We study the online detection of false data injection attacks and denial of service attacks in the smart grid. The system is modeled as a discrete-time linear dynamic system and state estimation is performed using the Kalman filter. The generalized CUSUM algorithm is employed for quickest detection of the cyber-attacks. Detectors are proposed in both centralized and distributed settings. The proposed detectors are robust to time-varying states, attacks, and set of attacked meters. In the distributed setting, due to bandwidth constraints, local centers can only transmit quantized messages to the global center, and a novel event-based sampling scheme called level-crossing sampling with hysteresis is proposed that is shown to exhibit significant advantages compared with the conventional uniform-in-time sampling scheme.

    Collaborators: Xiaodong Wang (Columbia U, EE), Shang Li (AT&T Labs)
    Student: Mehmet Necip Kurt (Columbia U, EE)
    Representative papers: "Quickest Detection of False Data Injection Attack in Wide-Area Smart Grids", IEEE Transactions on Smart Grid, 2015

  • IoT Security:

    A set of recent successful Distributed Denial-of-Service (DDoS) attacks on the Internet, facilitated by the proliferation of the Internet-of- Things (IoT) powered botnets, shows that it is a matter when, not if, that the Smart Grid becomes the target and likely victim of such an attack, potentially leaving catastrophic outage of power service to millions of people. Under a hierarchical data collection infrastructure we propose a general and scalable mitigation approach including online algorithms with low computational complexity, fast detection, and distributed statistical inference, referred to as Minimally Invasive Attack Mitigation via Detection Isolation and Localization (MIAMI-DIL).

    Collaborators: Suleyman Uludag (U Michigan Flint, Computer Science)

  • Computational Criminology:

    Traditionally, in the criminology literature, regression analysis is performed using data from the physical space only, such as location, demographics, and wealth. However, with the widespread use of technology, especially the internet, in daily life today, data from the cyber space also starts playing an important role in analyzing crimes. Analyzing cyber-data requires advanced data mining techniques to deal with the vast amounts of streaming disparate data types with high uncertainty, also known as big data analytics.

    Collaborators: Morris Chang (USF, EE), Kevin Wang (USF, Criminology)

Wireless Communications
  • Cognitive Radio:

    Addressing the well-known problem of spectrum utilization scarcity in current wireless networks, the cognitive radio (CR) technology employs a hierarchical spectrum access model consisting of primary users (PUs) and secondary users (SUs). In this model, both PUs and SUs are able to access a same band with a higher priority for PUs. The spectrum sharing between PUs and SUs can be realized in an underlay fashion, which allows SUs to coexist with PUs without sensing the spectrum band. Thus, SUs are blind to the idle state of PUs (spectrum holes), resulting in a worst-case assumption that PUs use the band all the time. As a result, SUs can coexist only with severe constraints on the transmission power in order to protect the quality of service (QoS) of PUs.

    Students: Abdullah Taher (USF EE), Farzad Shahabi (USF, EE)
    Representative papers: "Sequential Joint Detection and Estimation: Optimum Tests and Applications", IEEE Transactions on Signal Processing, 2016
    "Sequential Joint Spectrum Sensing and Channel Estimation for Dynamic Spectrum Access", IEEE Journal on Selected Areas in Communications, 2014
    "Cooperative Sequential Spectrum Sensing Based on Level-triggered Sampling", IEEE Transactions on Signal Processing, 2012

  • mmWave Communications:

    The small wavelength at millimeter wave (mmWave) allows to pack antenna arrays in a very small area enabling practical large-scale MIMO. However, wideband and high-precision analog-to-digital converters (ADCs) are very expensive and power-hungry. In this work, we propose a mmWave MIMO communication system that employs simple one-bit ADCs at the receiver and MIMO precoding at the transmitter. One significant challenge associated with this system is efficient channel estimation based on one-bit quantized channel output.

    Collaborators: Xiaodong Wang (Columbia U, EE), Ziyu Guo (Huawei, China)
    Students: AlMuthanna Nassar (USF, EE), Abdullah Taher (USF EE)
    Representative papers: "Transmitter-Centric Channel Estimation and Low-PAPR Precoding for Millimeter-Wave MIMO Systems", IEEE Transactions on Communications, 2016