DIVISION OF ENGINEERING AND APPLIED SCIENCES
HARVARD UNIVERSITY
CS 263. Modern Distributed
Systems:
|
| Research Project Ideas |
These are only suggestions to help you get a sense of the scope and topics for research projects that would work for CS263. As mentioned previously, projects must have some connection to the overall topic of the course, but can draw on ideas from other fields (e.g., theory, AI, languages, etc.) In fact, we encourage projects that have a "non-systems" component.
Projects are to be undertaken in groups of two students, unless you have made special arrangements with me. Need a project partner? Post to the class bulletin board.
Proposal Format Research Project Proposals due by beginning of lecture on Thursday, October 28. Please email your proposal in PDF format to mdw@eecs.
The proposal should be a 2-3 page document including sections on:
- A summary of the project;
- Background and related work (specifically, describe what is novel about your project);
- A brief description of your proposed approach, and any other thoughts on how you will proceed;
- A specific timeline of milestones that you intend to accomplish for your project. This should include the initial starting point, goal for what you intend to accomplish by the project update (due November 23), and final goal for the end of the project (final project report due Friday, January 7, 2005).
Project Ideas
- Develop an interface from a sensor network to an Internet-based stream processing system.
A number of projects (e.g., Hourglass, Medusa, IrisNet, and PIER), are looking at building Internet-based stream processing engines using overlay networks. The idea is to use these systems to process real time data from sensor networks. However, most of these systems don't actually connect to a real sensor network -- they simulate the sensors.
The goal of this project is to interface one of these systems (such as Harvard's own Hourglass) with a real sensor network (such as MoteLab). This involves developing a kind of proxy that will allow Hourglass to discover the sensor network attributes and submit queries to the system (e.g., using a system such as TinyDB), the results of which are then delivered to the stream processing system. This is an exciting project and if successful will tie in with other research efforts at Harvard.
Using a Web Services interface (e.g., through .NET) would provide a generic interface to the sensor network through the "Grid".
- An Adaptive Compression Scheme for Wireless Medical Sensors
The Harvard CodeBlue project is developing mote-based medical sensors for use in hospital and disaster response settings. However, given the limited radio bandwidth on these devices, it is not feasible to transmit data from all sensors continunously. One solution is to have nodes determine when there is radio congestion and to adaptively compress or filter their data before transmission. For example, an EKG signal can be compressed with a range of lossless, lossy, or progressive compression schemes that would allow the quality of the received data to scale with available channel capacity. In cases where congestion is severe, receiving a summary of a patient's status is preferable to receiving nothing at all.
An ideal project will look at the specific requirements for medical sensors (we can give you some assistance with this) and develop algorithms and protocols that are tailored for this environment.
- 802.15.4 benchmarking with motelab
Define several traffic patterns to challenge the network. Measure the link quality, packet loss rate, highest data rate, etc. Analyze the result with respect to location of the motes, number of concurrent transmissions, time in the day and so on.
Compare the results with existing models of radio propagation in indoor environments. Consider the use of multiple frequencies and transmit power levels (the CC2420 radios allow you to control frequency and transmit power in software).
For this project, be sure to read the paper on SCALE by A. Cerpa et al. (UCLA).
- Implementation and measurement of ad hoc routing protocols on sensor network nodes.
Several of the "classic" ad hoc routing protocols (AODV, DSR, etc.) have been implemented on TinyOS, although I am not aware of any serious evaluations of their performance. Use MoteLab to implement and study one or more of these protocols in a realistic indoor environment. Use your mote kits to measure performance with mobile nodes roaming the building. Compare your result with published simulation and other studies using other devices (e.g., 802.11 cards).
- Design a Compact Flash version of the Telos mote for use with PDAs.
This project requires circuit and PCB design experience, but is probably not too difficult. For various projects involving integration of PDAs with sensor networks, we would very much like to have a Compact Flash or SD version of the Telos that we could plug into the PDA directly. This would allow the PDA to relay data to and from the sensor network, opening up a lot of interesting applications.
In addition to designing the hardware we would need some simple software to talk to the mote over the CF or SD interface, such as through a simple UART.
If you successfully design the board we will find resources to build the actual hardware itself!
- Extend the TOSSIM environment for realistic radio models, Telos, and 802.15.4.
The TOSSIM simulator is widely used to study sensor network applications, though it isn't very realistic. Extending TOSSIM to integrate a better model of radio propagation, perhaps based on real measurements using the MoteLab testbed, would be incredibly valuable.
One aspect of this project would be to take a set of measurements of radio connectivity in MoteLab and produce a radio model for TOSSIM based on this data. Then, run an application (such as TinyDB, or the routing protocol you developed as part of Assignment #1) and compare the performance in simulation to real data on MoteLab. If the two do not match up well, try to understand why -- what other aspects of TOSSIM need improvement to get it to be more realistic?
Newer versions of TOSSIM include extensions to model power consumption of individual motes. This model is based on the older Mica2 nodes and needs to be updated for the MicaZ and Telos motes. The sensor models supported by TOSSIM are also very simplistic and need to be updated.
The end result of the project should strive to demonstrate a set of real applications running on both MoteLab and TOSSIM that behave identically (or as closely as possible). Identifying problems with the existing simulator is also instructive to the research community.
See the papers on TOSSIM by Levis et al. in SenSys'03, and PowerTOSSIM by Shnayder et al. in SenSys'04.