About us....
Computers can be programmed to observe their surroundings and operate effectors that act in the real world. Such programs are at the core of a large number of devices from simple household thermostats to complex autonomous robots and advanced machine control. Traditional engineering methods obligate the programmer to pre-determine the exact nature of what the program can sense, how it converts sensory information to observations, what actions are possible, and how those actions are related to observations. An unfortunate, but largely unavoidable side-effect of traditional methods is that both what the machine can know and what it can do are entirely pre-determined by the programmer. Such programs will be unlikely able to exploit unexpected opportunities not implicit in the pre-determined conceptualization of the world. Likewise, they will be unlikely able to properly deal with unexpected obstacles or internal damage.
In the Computational Autonomy Research Lab, we study methods that enable computational machines to be able to determine, for themselves, what actions they can accomplish and what observations they need to make. Such machines, it is believed, will be better able to do what is needed as it is needed without pre-conceived limitation. Our work touches on a large number of traditional areas of engineering and computer science. Some of these areas include artificial intelligence, automated pattern recognition, VLSI system design, autonomous robotics, and adaptive control. The common thread through all the work is the attempt to allow machines to, within the limits of safety and economy, transcend pre-designed limits.
This page contains descriptions of individual research efforts and is a dissemination point for some of our materials. Please feel free to contact us with any questions or comments. Prospective students are asked to read the materials in the Joining Us link prior to making initial contact.