Platform

Mobile Actuator Sensor Networks

Test-Bed

The MAS-net project is still in initial stage. We plan to model the 2-D air diffusion process by MAS-motes first. My work focuses on building the experiment test-bed for the project. A platform is built as a huge, thin, and black container so that white fog can flow inside the platform. The container is covered by three large transparent acrylic boards. Ten Mica2-based MAS-motes are built to move on the platform and observe the fog diffusion underneath. An overhead camera is used to localize these MAS-motes and help the precision of their movement.

Robots

The individual robots, we label MAS-motes, are built based on MicaZ motes developed by CrossBow Technology Inc. The MicaZ motes integrates an ATmega 128L as its central processor. MicaZ motes can communicate with the base station PC via a programming board that connects to the PC's serial port. A MicaZ mote programmed to be the gateway between the RF communication and the serial port is mounted on the programming board. The MicaZ mote passes everything it receives from wireless communication to the PC and vice versa.

Although this is just an experimental platform, we want the robots as close to the actual requirements as possible. This means our robots should be cheap, low maintenance requirement, robust and small so that they can be deployed in harsh environment. The MAS-motes are two-wheel differentially steered. They are all equipped with 3 IR sensors, 2 up front and 1 on rear for collision avoidance and also 2 photo resistors facing down to detect the concentration of fog underneath the MAS-mote.

Vision-Based Localization

The system localize each MAS-mote on the platform by a marker on top of the MAS-mote. The marker features a rectangular frame and a special symbol pattern inside the frame. The system first detects the frame and then tries to match the inside pattern with its pattern repository. The id, position and orientation of each marker will be obtained. This localization subsystem is based on modified ARToolKit. The HSB color model is adopted for the color segmentation so the effect of illumination is reduced and any specified color can be used for markers.

The image processing can work in multi-threading and pipeline fashion. The process includes lens distortion compensation, markers detection, and rendering to screen but the performance can be as high as 150 ms per frame. The transformation between the image coordinates and the world coordinates is implemented by bilinear approximation. The average errors of position and orientation are about 1.7 cm and 1.2 degree, respectively. The result of the localization is broadcasted to all MAS-motes so the MAS-motes can update their own estimated position and orientation.

Graphical User Interface

The graphic user interface is designed just like a commercial application. The pictures captured by the overhead camera is rendered in the GUI window. The detected markers and other information are displayed on the picture with some degree of transparency. An AOE-like user interface is used. Users can control detected MAS-motes and query their information by mouse clinking on the GUI.

Home

Platform

Pictures

Movies

Publications

Team members

What’s new

In the news