15-446 Projects

These are just rough sketches of ideas. Feel free to email me to clarify. Also, not all the projects are really the same size/challenge. As part of the proposal, you should "resize" the project to what you think you can get done by the end of April. I will provide feedback on whether you need to add/delete stuff from your proposal.

Collaboration Applications

• Shared event recording
• Class note taking. Implement a shared note taking system for classes. Take a look at http://www.research.ibm.com/journal/sj/384/landay.html for some ideas.
• Photos. Perhaps linked to Photosynth but have a time aspect to the view
• Audio. How do we merge the audio clips and determine their relative timing/content.
• Classroom tools
• Polls, Q&A. Build tools that allow instructors to take quick polls or perform Q&A with the students.
• Collaborative editing
• Whiteboard, documents. Build a collaborative editing tool with various consistency models (depending on the type of document/chosen by the user)
• Distributed IM. Can we build an IM system that makes use of local communication. Perhaps something that leverges geocast (tell people in this area)
• Distributed Pubsub. Build a publish-subscribe system for Android. Use gossip to propagate updates between phones. You could use this to share photos, songs or other content based on specific attributes.
• Reviewing/reputation designs. Build a reputation system that lets users rate nearby objects. Key issue might be maintaining privacy (I have a paper design that I can provide). You could rate photos, files, songs, 802.11 access points, restaurants and other services

Games

• Strategy games. Build a slowly played game where moves are exchanged opportunistically via gossip.
• GPS-based games. Build a location-based game with real-time updates
• Standard game. Build a networked game with a mobile client.

Application Distribution

• Get updates and find new applications in a P2P fashion. We currently get app updated only via the app store. Can we switch this to using the bittorrent from the project below?
• P2P filesharing for mobiles. Build a bittorrent client for the phones that takes advantage of both cellular and local connectivity.
• Control running code/experiments on distributed nodes. We would love to go to Google and get 1000+ phones for campus. How would we manage these phones. Could we send commands out to all of them quickly, perhaps use some gossip protocol. What if we wanted to collect data from them or run experiments?
• Framework for P2P routing of messages through phones. It would be nice for a wide range of applications to have a simple, lightweight, multi-hop routing system for phones. Can we take advantage of GPS position to help routing or to decide whether multi-hop is worth it? This could be used for VOIP or other apps.
• DOT client for phone. Can you carry content or partial content with you and reintegrate with downloads. See the http://www.cs.cmu.edu/~dot-project/

Social Network

• Locating friends via multi-hop communications. Can you find your friends nearby? Without revealing who your friends are? Take a look at this paper on how to find friends in email (http://www.cs.princeton.edu/~mfreed/docs/re-nsdi06.pdf)
• Implement SybilGuard. SybilGuard is a system that uses social networks to prevent sybil attacks. See http://conferences.sigcomm.org/sigcomm/2006/discussion/showpaper.php?paper_id=26
• Shared "bulletin boards/post-its". What if we had monitors all over the new Gates building. Could we leave post-it notes for our friends or have customized bullitin boards?
• Trajectory recording. How about leaving information about where we have been recently as a post-it? How do we update/collect trajectory info?

Sensor Networks

• How many people near you? Can you identify the number of people running your application with a few wireless hops? A neat idea might be to use some of the synopsis diffusion sensor query techniques (http://www.cs.cmu.edu/~sknath/sd.htm). You could integrate this with maps to provide "how busy are certain places" or "how slow is the traffic".
• Image/audio sensing. Could you take audio recordings from multiple locations and merge to make a better version (e.g., remove background noise?) Or use images as part of Photosynth from MSR (http://livelabs.com/photosynth/).
• User-entered sensor feeds. Things like prices and weather.
• Incremental localization. GPS works well outdoors. Can you use wireless multi-hop to reach a set of people nearby with GPS coordinates and then infer your own. See this paper
• Intel's Fast Food Cam. We are looking for a team to develop a client/server architecture in which a user capture images of fast food with the camera on their phone and the images are shipped to a server (ideally running on Intel's cluster) which will run image processing algorithms to recognize the food and estimate its nutritional content. The server will then ship back to the phone either a digest of the nutritional info, or the original image with the nutritional information overlaid on it, so the user can find out in nearly real-time what they are eating and how many calories it contains. The client-side application would ideally be location-aware, so as to provide helpful clues to the server-side food recognition algorithm about the restaurant the user is visiting. An ambitious team might also investigate the user interface best suited to presenting the nutritional information, and for allowing the user to interactively narrow the set of choices to improve the accuracy of the food recognition system.
  Intel has a good handle on the image processing aspects of the project, but are really looking for help with the client (phone) side of the system. If the project is successful, there is definitely the opportunity for the project to continue through the summer (as a paid summer fellowship at Intel) and into next year.

Last updated: Thu Apr 09 13:21:13 -0400 2009 [validate xhtml]