Robots achieve their functionality by executing purposeful motions in the real world. Intelligent robots can be characterized by the ability to autonomously plan and execute motion sequences to achieve a goal specified by a human user without detailed instructions. This is in contrast to most current industrial robots that have to be pre-programmed with very little adaptivity in their task execution. Intelligent robots can also be characterized by the ability to operate in an uncertain, changing environment with the help of appropriate sensing, for example, computer vision.
This course will examine some of the fundamental geometric and computational issues involved in robot motion planning, sensing, and control. Topics will include representation and handling of uncertainty for motion planning/sensing/control, computational complexity of different motion planning problems, approximation techniques for practical motion planning, sensor-based motion planning, sensor-based control, active vision, and robot learning. It will also briefly examine the different computational architectures proposed for the implementation and evolution of intelligent robotic behavior.
The course material will be drawn from several books, tutorial papers, and a few recent research papers. No background in robotics is needed. The material in the course will be largely self-contained. The lectures will include a brief introduction to basic robotics, including robot kinematics. Short tutorials of mathematical techniques will be provided as needed.