Wireless Sensor Probe Networks - SensorSimII

Craig Ulmer/ grimace@ee.gatech.edu

Back

This is a Work In Progress! Please be patient.

About

SensorSimII is the framework of a simulator that I have been working on to study how future sensor networks should operate. the simulator is written in a modular fashion so that it can be adapted to serve a number of needs. However, please remember that it is still a work in progress. This web page is here just to give a glimpse of the approach we've taken with this simulator. Likewise this web page is simply preliminary information to attempt to answer some of the questions that researchers might have about this project.

Presentation

Source Code

Simulator

The simulator can be started by clicking the following link:

(A button should appear above this line momentarily)

Basic Information

There are a number of assumptions in this simulator model:

• Encapsulated State: Each node in the simulator runs its own state machine, and is only aware of events that it realistically can observe.
• Message Passing: Nodes relay information by transmitting messages. Therefore change's in a node's state are due to either the reception and reaction of incoming messages, or timeout values in the state machine.
• Broadcast Transmissions: This simulator assumes a simple proximity based, broadcast transmission model where all nodes within a fixed radius can observe a transmission. If a node receives two transmissions at the same time, both messages are corrupted and discarded.
• MAC Protocols: To facilitate link-level reliability, node's have a MAC unit which uses control messages to make sure messages are reliably transmitted. Currently we use a RTS/CTS/DATA/ACK scheme. Combined with the physical layer's simple carrier sensing mechanisms, this scheme helps prevent collisions since the MAC layer delays transmissions if another node's reliable transmission dialog is observed to be in progress.
• Limited Ejection Points: We believe that sensor networks for operation in remote areas will have few ejection points (ie, base stations) that will be randomly located in the network. These nodes are marked with a white X in the simulator and will likely be the ultimate destination of a large percentage of messages.

Simulation: Simple Clustering

A simple clustering algorithm has been programmed into the simulator for the above demonstration. In this algorithm, a number of nodes are defined at start time to be cluster leaders based only on their ID numbers. As nodes wake up, they transmit information to their neighbors as to how far away the nearest leader is (ie, distance vector fashion). Nodes join a leader's cluster simply based on selecting the nearest leader. In order to deal with the fact that leaders will probably not be in ideal, equally distributed locations, the algorithm contains a mutiny clause. Should a node decide that it is too far away from its nearest leader, it can potentially start a mutiny where it elects itself as a new leader. Additional information is used to determine whether a node should start a mutiny (ie, whether the node is the farthest away) in order to decrease the number of mutinies in a region.