The research project will use the sensor technology of the Microsoft Kinect to detect humans and other objects. The long-term goal of this project it to integrate the sensor on iRobots Creates. These robots are already use for caravanning (an application designed by members of our lab for distributed group coordination amongst heterogeneous robots), and a more powerful set of sensors would allow the robots to follow humans and better detect objects in the path of the robots.

     Using the Kinect will allow my fellow researchers to use it instead of the basic webcams to detect objects. Using the webcams doesn't really give the robot a sense of depth or direction when coming in contact with an obstacle or person. The robots currently use webcams to read diagrams for information. While this is good for a start, it is not always accurate. For instance, while the robot is in motion, a person can walk in its way: without a diagram attached to the person's leg, the robot will run into the persons leg. The Kinect will allow the robot to detect the person and stop to avoid a collision.

     The motion planning problem is, given the start of goal configuration of a robot in an environment, to find a valid path from the start to the goal. The motion planning problem can be solved by many different algorithms, including road-map based methods like PRM or tree based methods like RRT. A benchmark is a standard against which related things may be compared; in motion planning, benchmarks can be used to compare the efficiency and performance of different algorithms. The Parasol Laboratory motion planning benchmark suite includes narrow passage, dis-assembly, surface, and folding problems. During the course of the this summer, I improved and organized the Parasol Motion Planning benchmark suite. I also used the benchmarks to evaluate several motion planning algorithms, comparing the overall time, collision detection calls, and shortest path generated in a given environment.