Carnegie Mellon Robotics Institute
My research interests center on developing highly capable systems and machines that are both easy to instruct and capable of dealing with uncertainty. I am interested in both practical robot systems for manufacturing as well as purely research mechanisms. Whatever the problem or application, I have a strong interest in both the development and application of new fundamental theories as well as the physical demonstration of their utility. My current interests include:
Application of Hybrid Control to Robotic Systems. Recent advances in theory and computational power have made it possible to implement a multitude of modern advanced control strategies on a robotic system. However, the problem of how to safely schedule and sequence these strategies to accomplish all but the most simple task remains daunting, and represents a significant barrier to the development of truly capable machines. While many researchers throughout the world have embraced the notion of utilizing sensor measurements (feedback) to guide such decisions few have done so in a systematic and structured way. We are striving to develop an abstract ``systems-theory'' to guide the development of such strategies, the goal being to ensure that the abstract composition of simple strategies that are known to work will result in a more complex system with known performance properties and guarantees.
Robotic Systems Capable of Dynamic Manipulation. Advances in artificial intelligence and machine learning over the past decade have resulted in machines capable of reasoning about complex domain specific problems as well if not better than the best humans; take for example IBM's chess playing machine. However, we have failed to produce any machines capable of dealing with the ambiguities and dynamics of the real world as well as the average human toddler. Towards this end we are interested in the design, development, analysis, and construction of systems capable of dynamically manipulating their environment. Examples of such tasks include juggling, ball bouncing, catching and throwing. The desire is to develop robust algorithms that both enable the execution of such demonstration tasks while providing insight into potential solutions for more generic tasks, such as picking up a glass of milk or cleaning a room.
Dynamic Robot Locomotion. Much as we lack understanding about how to construct systems to manipulate the everyday world, we also lack the ability to build systems that can locomote reliably in that same world. The algorithmic and theoretical problems are similar to those of dynamic manipulation, however they are compounded by the need to build highly efficient, compact, and energy autonomous systems. Addressing these issues necessitates a thorough understanding of the tradeoffs and interactions between the multitude of components used to construct the mechanism as well as the algorithms used to guide and control it.
|Research Interest Keywords|
|actuators, assembly, control, factory and warehouse automation, legged locomotion, manipulation, manufacturing, mechatronics, multi-agent systems|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions