Dynamic sensor planning

Abstract
A method of extending the sensor planning abilities of the MVP (machine vision planning) system to plan viewpoints for monitoring a pre-planned robot task is described. The dynamic sensor planning system presented analyzes geometric models of the environment and of the planned motions of the robot, as well as optical models of the vision sensor. Using a combination of swept volumes and a temporal interval search technique, it computes a series of viewpoints, each of which provides a valid viewpoint for a different interval of the task. By mounting a camera on another manipulator, the viewpoints can be executed at appropriate times during the task so that there is always a robust view suitable for monitoring the task. Experimental results monitoring a simulated robot operation are presented, and directions for future research are discussed.<>

This publication has 7 references indexed in Scilit: