|
|
|
|
|
|
|
Research Projects
Description of Projects for Summer 2005 Parking lot occupancy using wireless sensor networks and cellular telephones An End-to-End Customizable Tool for Available Bandwidth Estimation
Bandwidth Estimation techniques in the optimization of BitTorrent
average download time ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Developing Timing Verification
Algorithms There are three components needed for model checking: a design under verification (DUV), an environment defining how the inputs of DUV are driven, and a correctness specification defining the correct behavior of a system. Then, they are fed to a model checker that will generate an answer either Yes or No. If answer is No, then a sequence of states is generated to show why the specification has failed. The idea is illustrated in the following figure.
In this project, a student will be involved in a verification system development. The student will study the fundamental concepts of verification and model checking, and will be implementing verification algorithms for some specification logic. The student will have the chance to conduct some research on new verification algorithms. The qualifications of the student include good C/C++ programming capability, knowledge of common data structures and algorithms, and logic design.
Natural disaster/hurricane application
Traveling Assistant Device This proposed application, Travel Assistance Device (TAD), would use multimedia cell phones with built-in global positioning systems to help new transit riders, especially those who are cognitively disabled. Informational prompts will be delivered to the rider in a “just-in-time” method such as triggering the phone to ring, vibrate or deliver an audio message when the rider should pull the stop cord. Automated alarms can be triggered and the travel trainer and/or parent/guardian remotely alerted in case a rider wanders off their pre-determined path. Traditional phone communication is possible between the rider and the trainer which will allow them to guide the rider to the correct location if they are lost. This device could increase fixed route transit use among the cognitively disabled. Intrusion detection using wireless sensor networks Project Supervisor - Dr. Miguel A. Labrador, Assistant Professor, Dept. of Computer Science and Engineering, USF. Wireless Sensor Networks are self-configurable wireless networks made of small, cheap and resource constrained devices used to transmit information from places where human intervention is not possible, desired or very risky. WSNs can be utilized in a countless number of applications such as environmental monitoring, security, medicine, tracking of assets, intrusion detection, home automation, parking and transportation, etc. In this project, students will utilize the necessary software and hardware to design and implement a small-scale application for intrusion detection. A small scale, proof of concept WSN with wireless sensor devices with different sensing capabilities will be used to monitor a particular area for intrusion. One common example of a large scale intrusion detection application would be the monitoring of border lines between two countries. Wireless sensors devices will be placed at specific places in the area of coverage to detect violations to the covered space. Wireless sensor devices will transmit space violations detected by the sensors through the WSN to the management station (application), where this information will be visualized. We will use the wireless sensor hardware and software from Xbow, Inc. The students will investigate the detection algorithms, the coverage, detection success rate, etc. to make the application reliable.
Parking lot occupancy
application using wireless sensor networks and cellular
telephones The students will investigate the detection algorithms, the coverage, detection success rate, etc. to make the application reliable.
Packet loss distributions of
TCP using WEB100 A set of experiments will monitor packet loss events and measure the change in rates given by the TCP congestion control mechanism. These results will be used in the design of new bandwidth estimation techniques for transport layer protocols.
Students in this project will learn about TCP flow and congestion control, packet monitoring, and Linux and freeBSD network configuration. Development of a Peer to Peer Proxy Peer to 2 Peer (P2P) requires a PC to be fully powered on at all time to respond to (and forward) query messages. This project proposes to build a P2P proxy that knows what files are shared. The proxy can reply to queries and will wake-up the actual host only when a download request occurs. Saving energy is the purpose here. Students will learn the protocol internals of the Gnutella P2P system and socket programming. Students working in this project need to know how to program using the C language. Socket programming is a plus but not required.
Scalable Data Mining Algorithms The main thrust of work in the Intelligent Systems Laboratory is to develop scalable data mining algorithms. In this project REU students will be involved in graph data mining which involves learning to identify sub-graphs of importance. Many graph searched problems are not computable. So, our approach is to find graph Kernels that are capable of comparing graphs and to utilize multiple kernels to get a good picture of similarities. The problems being undertaken would have up to a terabyte of labeled or unlabeled data.
This project will involve REU students in the design and implementation of a large-scale (128 node) compute cluster. This system will be designed and built from component parts. There will be hands-on hardware configuration (installation of the CPU, power supply cooling fans, interconnection of processors, etc.) as well as operating system installation and the application and analysis of benchmark software programs to test the functionality of the computer.
There are several opportunities for REU students interested in this area. One project has to do with face recognition using dynamic strain patters from video. Strain patterns exhibited during facial expressions provide a unique fingerprint for recognitions. Another project implements algorithms to detect power lines in bad weather, cluttered scenes. This project is based on army reports saying that more helicopters have been lost because of power lines than against enemies in combat. The third project is meant to automate the analysis of brain MR images, segmenting and labeling the tissues of interest, such as cerebrospinal fluid, white matter and grey matter.
An End-to-End Customizable Tool for Available Bandwidth Estimation The interest on the measurement of the end-to-end available bandwidth estimation (avail-bw) has generated a variety of software tools that have been developed, for the most part, during the last 5 years. These tools perform according with specified parameters (i.e. number and length of the probing packets) that in most of the cases are in-code specified. Different applications demand specific values of accuracy, speed and overhead in the avail-bw estimation. This project is to deploy a new customizable tool to perform end-to-end avail-bw measurements based on application/user specified parameters. The tool uses a train of packet pairs to do the estimation, based on a client/server socket architecture. Students in this project will mainly learn and apply C programming, sockets, linux and freeBSD network configuration. The tool will be tested in a existent WAN emulator
Bandwidth Estimation techniques in the optimization of BitTorrent average download time Peer to peer distributed networks allow users across the Internet to share information of different type and in different ways. Collaborative file sharing protocols are limited by bandwidth constraints. We recently have explored this problem in the optimization of BitTorrent average download time using IGI to as an estimator of available bandwidth. This estimation is used by the protocol to select peers with similar bandwidth capabilities. The goal of this project is to include pathload in the BitTorrent protocol to estimate available bandwidth and to evaluate the performance improvements between the previous work (original protocol and the IGI modification) and the application of pathload. Students in this project will mainly learn and apply Python and C programming, sockets, modelnet, linux Debian and freeBSD network configuration. The tool will be tested in a existent modelnet emulator.
|
