         
|
GRASP Seminar Series: Spring 2005
April 22, 11:00 AM, Levine Hall 307.
Seth Teller
MIT EECS Dept. and CS & AI Laboratory
“Pervasive Multi-Sensor Egomotion Estimation for Direct Interaction and Unstructured Robotics”
Abstract: For humans, knowledge of our own location is a basic kind of empowering information: as part of our mental model of the world, it enables us to navigate to desired places, to find resources, and to plan our movements more effectively. Until recently, people had to rely on experience and continuity to locate themselves, their assets and devices. In recent decades, however, position information from the Global Positioning System (GPS) infrastructure has wrought tremendous change in human and robotic activities outdoors, ranging from military operations (including autonomous aircraft), civilian navigation and surveying, to shipping and supply-chain management, resource exploration, and precision agriculture. We envision an analogous indoor infrastructure, combining active radio and ultrasound beacons with passive receivers using machine vision, to provide fine-grained location and orientation ("pose" or "egomotion") information to human-held devices, autonomous robots, and ordinary objects. This infrastructure has the potential to substantially extend and improve human interfaces and robotic capabilities. For people, pose-awareness facilitates direct interaction with things in the world and their metadata. For robots, pose-awareness makes feasible tasks that are currently out of reach, such as complex household chores. After motivating the infrastructure, we'll show some early deployments and proof-of-concept applications, and briefly discuss privacy concerns. We'll also show some early efforts toward making the infrastructure deploy itself autonomously.
Biography: Seth Teller is a member of the EECS Department and Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. His research combines computer graphics, machine vision, computational geometry, robotics, sensor networks and pervasive computing.
