Acquiring Medium Models for Sensing Performance Estimation

The quality of sensing in practical sensor network deployments suffers due to the presence of obstacles in the sensing medium. If such unknown obstacles are present, and the sensor data indicates that no targets of interest are detected, then there is no easy way for the application to distinguish between the cases that there really is no target or that the targets are located in occluded regions. The obstacles may not be known before deployment and may change over time. Hence, it is of interest to develop methods which enable a sensor network to determine the presence and extent of sensing occlusions. We present one such method based on the use of a range sensor to map the obstacles in the medium. A network architecture to support efficient medium mapping facilities is presented, along with several design choices in the acquisition and update of the medium map data. We also present algorithms to rapidly acquire this data and share it among multiple nodes. All algorithms presented are implemented on prototype hardware consisting of an actuated laser and an embedded processing platform.