2460 A.V. Williams Building
For More Information:
Techno-Sciences Inc. Robotics Seminar Series
Why drive (autonomously) when you can fly (autonomously)?
Carnegie Mellon University
In the last two decades autonomous robots with legs and wheels have moved cautiously from simple indoor environments to the more complicated outdoors. Some robots now drive at highway speeds while others negotiate rough terrain even on distant planets with minimal supervision. In contrast, autonomous air vehicles have been constrained to fly high where they are very unlikely to encounter obstacles, or to fly under careful manual supervision when close to the ground. In this talk I will discuss an agenda for intelligent air vehicles that navigate, explore and collaborate in and around significant three dimensional structures. Fundamental to intelligent navigation is the ability to get from A to B, without prior knowledge of the environment, sensing and avoiding obstacles in the way. I will describe intelligent control for an autonomous helicopter that has evolved from autonomous ground vehicles in development at Carnegie Mellon since the early 1980s. Continuing the philosophy of "plan globally, react locally", we have developed a tiered architecture that simplifies the need to consider geometry and dynamics simultaneously. The highest level plans paths considering geometry of the environment, a middle layer controls steering incorporating automatically learned vehicle dynamics, and the lowest level modulates speed using a forward closed-loop model of the control. While the logistics and safety issues associated with autonomous flight are complicated, I will show why in many cases, "slipping the surly bonds of earth" simplifies the problem of intelligent navigation. I will show results from series of experiments that have demonstrated autonomous flight close to the ground, between wires, trees and buildings on platforms spanning submeter scale "quadrotors" to full-sized helicopters. I will also try to answer the question what it will take to get to the Jetsonsí age in which it will sometimes make sense to fly your car autonomously rather than to drive it.
Sanjiv Singh is a Research Professor at the Robotics Institute, Carnegie Mellon University. His recent work has two main themes: perception in natural environments and multi-agent coordination. He has led projects in both ground and air vehicles operating in unknown or partially known environments, in applications such as mining, agriculture, emergency response, surveillance and exploration. He has recently led projects that have demonstrated autonomous navigation for a full scale helicopter and have developed autonomy for a flying car. Prof Singh received his PhD in Robotics from Carnegie Mellon (1995). He is the Editor-in-Chief of the Journal of Field Robotics.
This Event is For: Graduate • Undergraduate • Faculty • Post-Docs • Alumni • Corporate