MIT's OLPC Pushes Bounds of Mesh Networking

by Ash Dyer

I just dropped by the project showcase for MIT's MAS.964 (graduate course for One Laptop Per Child). I was very impressed with the applications the students, professors, and lecturers are developing for the OLPC. In particular, I was struck by the work that's been done on Mesh Networking as part of the MIT Media Lab's work with the OLPC project, especially Michail Bletsas and the venerable Walter Bender.

Polychronis Ypodimatopoulos, Pol for short, has developed a "probabilistic presence mechanism" that interacts with the mesh firmware for the OLPC. The presence mechanism uses a very small amount of information from special presence frames exchanged between direct neighbors that contain condensed presence information from their respective neighbors (the ad-hoc meshing firmware in the OLPC relays packets without touching the kernel - a novel development in its own right that reduces the CPU load on the processor to only the packets destined for that machine). The impressive feature of this approach to presence is that the time to detection is a linear function of the distance - a major change from most mesh networks where the routing table is either pre-determined or the time to discovery is a quadratic function (square) of the distance.

Pol's writeup on his work is here:

*please note this posting has been corrected based on feedback from Pol Y.


John Leuner
2007-07-13 07:33:20
That's not a writeup, it's a 1 page PDF