Robert is currently working on UAV indoor navigation.  The near term goal (summer 2010) for the group is to get a quadrotor flying waypoints using stereo vision odometry.   The group plans on eventually having a quadrotor be able to explore the MAGICC Lab autonomously using both a camera and laser scanner.

Bryce's research includes work on super-resolution and mosaicking, image registration, and visual odometry.  One way to visualize and utilize UAV video is to convert the noisy, jittery video stream from the UAV into a 2-D map of the terrain over which the craft is flying.  Pasting multiple video frames together to make this map is called mosaicking.  Because the same spot of ground is usually seen by several video frames, "super-resolution" techniques can be used to make a mosaic that has higher resolution (and hence is clearer and shows more detail) than any of the individual video frames.  Bryce has developed a mosaicking tool that can generate a super-resolved mosaic from a video stream.  Evan Anderson recently got this algorithm working with real UAV video.

In order to mosaick video frames together, we first need to know how those frames line up with each other, so that we can stitch them together seamlessly.  The process of finding out how two images line up with each other is called image registration.  A fortunate by-product of registration is that it allows the video stream to be used to very accurately determine the position and the attitude (pitch, roll, heading) of the UAV.  Bryce is currently working on exploiting this fact to improve UAV position and attitude estimates, a task known as visual odometry.  This will allow the super-resolved mosaic built from a video to be accurately geo-referenced