Speckled Computing is an emerging technology in which data will be sensed in minute (ultimately around 5mm3) semiconductor grains called Specks. Information will be extracted, exchanged, processed and acted upon in a wireless network of thousands of Specks, called a Specknet. This collaborative effort by the thousands of Specks is known as Speckled Computing. The evolution of Speckled Computing brings about new requirements and constraints for wireless communications exceeding those imposed by typical ad-hoc mobile networks (MANET) and sensor networks. Speckled computing introduces new challenges to designers of communication protocols as power, memory and computational resources are extremely limited. This limitation, coupled with the need for Specknets to be autonomous without base stations or any form of infrastructure, makes many existing MANET and sensor network communication protocols unsuitable for Speckled Computing. This talk would cover the below topics:

a) Introduction to Speckled Computing - What is it and what am I working on?

b) Current Speck Prototypes - Introducing "ProSpeckz" (Programmable Specks over Zigbee Radio, 30mm x 28mm in size). Examples on some of the demonstrators built using the prototypes.

c) Physical Layer Wireless Communications - Radio or Laser? Look at the effects the shape of wireless coverage offered by the two mediums has on communications and networking then explore possible pros and cons.

d) Medium Access Control (MAC) - A look into the requirements on the MAC for Speckled Computing. To achieve low power, duty cycling is used to trade latency with power consumption. However, typical duty-cycling would require synchronization which is problematic to maintain (another large research area) given size of the network and inaccurate oscillators. So is there a solution that doesn't require synchronization??? Some thoughts would be presented here.

e) Networking (Broadcasting) - Broadcast is often used in networks and requires intermediate nodes to rely messages to each other so that the message can be received by all nodes in a multi-hop network like a Specknet. However, not all nodes need to relay the broadcast message but finding the minimum dominating set without global knowledge is an extremely difficult problem. Simulation results from some ideas would be presented and discussed.

f) Networking (Routing) - Rather than broadcasting, certain situations requires messages to be passed between a source node and a destination node. The task of finding a route from the source to destination in a multi-hop network is the job of the routing algorithm. In a Specknet, the job of the routing algorithm is made difficult as memory is extremely limited (so is the power and processing capabilities). Given that we expect a Speck to have 2Kbytes of memory for communications, this would limit the size of the packet, the routing tables, the buffers and the addresses that can be used. A suggested solution using Speckzones (a hybrid proactive and reactive approach) would be presented and discussed in this part.

Home : Grad_seminar