Past Tuesday Group meetings

Tuesday, 23 April 2002

- The meeting was devoted to a talk by Doug De Couto about his work on the Grid protocol for ad-hoc wireless networks
¤ Discussed the motivation, the protocol as it currently stands, and simulation results
¤ Showed the current test beds on the 5th and 6th floors of LCS, and in East Cambridge -- a rooftop net
¤ Material on the protocol is found at http://www.pdos.lcs.mit.edu/grid/
Tuesday, 9 April 2002

- Jonathan and Andrew gave a demo involving 50-60 tinies running an interactive leader election; there was discussion of the next extensions of these operations
- Howie described his ideas for locations finding by individual tinies, listening for signals from units with known coordinates
- Paul Elliott talked a little about Alloy, with a demo of graphing of two example descriptions
- Paul also discussed and showed a bit about Stateflow; he had trouble getting it to work, and he like GME better
Tuesday, 26 March 2002

- Jonathan and Andrew demo some early work on a Tiny implementation
- Greg discusses RMPL and the Generic Modeling Environment from Vanderbilt
- Suggestion to test laying out lots of Tinies, start a leader election, and see what happens
- Jonathan suggests a guest lecture from Doug De Couto of the Distributed Algorithms group -- to be scheduled for meeting after next
Tuesday, 12 March 2002

- Introduction of new attendees
- Discussion of UML diagramming
- Discussion of Cooperative Cruise Control
Tuesday, 26 February 2002

- Demo of Howie's simulator of behavior of groups of Tinies
¤ Operations: create an array; elect leaders; flood; gradient
¤ The gradient operation approximantes the distances to origin processors
¤ It is a model of the Tiny processors (on top of the Tiny OS), operating via a message queue and a task queue
¤ The code structure is all as tasks; written in Common Lisp
- Account of the MoBIES meeting, from Bob Laddaga
¤ The integration appointee for our section is Vanderbilt
¤ We are committing to using Brian's tools
¤ Also, we will work with the model-checking tools (contact, John Rushby, SRI)
¤ [ Other details that I couldn't write down ]
¤ We should discuss Rushby and Rational Rose at a future meeting
Tuesday, 12 February 2002

- NEST and MoBIES PI meetings happened since last group meeting.
- Jonathan gave a detailed account of the NEST OEP training and the latest state of the devices.
- Howie described the three challenge problems.
¤ Vibration damping in a nose cone, emphasizing real-time response
¤ Controlling a gossamer sheet, 25m diameter, emphasizing scaling
¤ The Berkeley OEP, emphasizing resource impoverishment
- Schedule: there will be PI meetings in June and December '02, with reviews just after each; with a decision after December on the overall program.
- Issues in the OEP task group include group formation, location discovery, time synchronization, and execution uncertainty.
- Members of the task group are universities: Berkeley, Vanderbilt, Notre Dame, MIT, and Memphis.
Tuesday, 29 January 2002

- Howie, Brian, and Jonathan will be attending the NEST meeting.
- JRB and Andrew S showed a ranging demo on two pushpin development boards: the LEDs on a row light to indicate the distance. They are communicating over a wire, so both boards display ranging detected by either.
- The question was discussed of presenting the pushpins as opposed to tinies, at the meeting.
- Howie discussed the contracts' history and DARPA politics and funding.
Tuesday, 22 January 2002

- Announced the need to resync the meetings with the faculty lunch schedule. The next meeting will be next week, 29 Jan.
- JRB and Andrew Sutherland gave a brief demo of a simple test application running on a pushpin processor: the LEDs flash in a pattern, which changes when the button is pushed.
- Use and cost of the pushpins was discussed.
¤ The test board (equivalent hardware, but large enough to handle and with all chip pins brought out to test contacts) costs $100.
¤ Actual pushpin modules cost $20 for parts, but require assembly, a serious, probably costly extra. The IR units and RF comm units cost $10 each, so a full unit is ~$40 for parts.
¤ The issue of "processor agnosticism" was discussed, and how to put this clearly and gently in the presentation.
Tuesday, 8 January 2002

- JRB discussed his recent discoveries about the "pushpin computing" developed by Joshua Lifton at the Media Lab
¤ JRB feels the "pushpin modules" may be better than the Berkeley Tinies for our demo
¬ Based on the Cygnal 8051 processor, with 32K flash RAM, lots of features
¬ Modules are fabricated as ~15mm squares, and are pluggable and stackable
¬ Types of stackable subparts include 4-way IR radiator/sensors, RF communication
¤ Reasons to choose this processor over PIC include a good C compiler, a debugger, etc.
¤ There is a serial interface that taps right into the microprocessor for debugging
¤ There is an OS under development, also a communication/software distribution utility
¤ All the details are at Lifton's web site (requires Javascript)
- JRB has been tinkering with printed circuit design
- The new UROP, Andrew Sutherland, is working on a prototype
Tuesday, 11 December 2001

- mention of 30 January MoBIES PI meeting.
- JRB talked about his experience with IPACs; he has a UROP starting in January. He met with the Amorphous Computing group, and with Radhika.
- discussion of a "cricket" location finder/reporter, possibly using Tinies or Pics. How much memory each has is an issue.
- discussion of a 3-tier sensor/gather/relay model for the demo.
Tuesday, 27 November 2001

- Radhika Nagpal came to the meeting and gave a quick presentation of her recent work in Amorphous Computing. A general discussion followed of how this relates to MoBIES. She seemed intrigued with this application of the theory, as opposed to the biological orientation (both literal and metaphorical) of the Amorphous Computing Group.
Tuesday, 13 November 2001

- There will be a MoBIES Principal Investigators' meeting in Boca Raton, FL, on 30 January-2 February 2002. We need a vision statement to give to that meeting.
- Bob Laddaga said this will be a plenary session and will discuss the status of the pursuer-evader problem.
- How can you download software to a Tiny? Can it be done by radio?
- We need to cobble up a demo of Tinies communicating; perhaps having LEDs flashing, as a visible indicator of what is happening.
- We need a talk by an author of the Amorphous Computing papers, and perhaps also by a distributed computation student.
Tuesday, 30 October 2001

- Krzystof gave a report on his visit to the Tiny-producing group on his trip to Berkeley.
- Some facts on the current state of their development:
¤ Multihop communication is not yet reliable.
¤ Battery life is about 2 weeks, with no power management. (!)
¤ Radio range is 100', can be reduced to one foot.
¤ They use a base station.
¤ Effective communication speed is 800 bytes/sec.
¤ There is documentation on three websites ...
¤ Available sensors include for light, temperature, a microphone, an accelerometer; there is also an interface for a motion detector.
¤ There are various radio models; all talk to all units in range.
Tuesday, 16 October 2001

- Costas gave a description of his work, on which he is writing papers.
¤ He is putting together a web site.
¤ One product is DPL: Deductive Proof Language, to provide searches and proof checking.
¤ This would be incorporated into a compiler, for example. It provides partial certification via casewise proofs, per execution.
¤ He is also working on a Denotational Proof Language.
¤ This would provide a shortcut way to provide correctness confidence in compilations.
Tuesday, 2 October 2001

- Inaugural meeting.