Diffusion Mapping Simulations
Mapping navigable space is important for mobile robots and can also be a product in its own right, e.g., in the case of reconnaissance. The idea of Diffusion Mapping is to use teams of many small robots as a dynamically deploying, richly connected net of sensors, in order to map out navigable space. We call this diffusion mapping, as very simple random walk or diffusion control strategies are used for the individual team members. We maintain the global position and configuration of the evolving sensor net, using robot odometry and relative bearing measurements between the robots themselves. The traces of the robots through space as well as lines of sight between the robots at successive times are used to carve out the free space. We are currently building a large number of small robots in the BORG lab at Georgia Tech (http://borg.cc.gatech.edu) in order to, among other goals, validate diffusion mapping experimentally. However, the system already runs in simulation and will be used to test the algorithms and help determine the accuracy specifications of the bearing sensors that we will put on the robots. Under consideration are both IR and RF-based bearing and range sensors. In this 7001 Project you would run the simulations (in MATLAB) with different number of robots, different values for the sensor accuracy and range, and differnt control strategies. The result will us help decide in what sensors to design for our hardware platform. The project deliverable is a report on the simulation results with a insightful discussion.
|