Communications of the ACM
Researcher to Help Develop 'Human-Like' Control System
The project's learning control structure will be integrated into space robots like the Mars Rover for navigation, exploration, and scouting applications.
Machines that can "think like a human" would be able to explore extreme environments or respond quickly to an emergency. But engineers must solve some fundamental scientific problems before machines can achieve these capabilities.
Assistant Professor Zhen Ni of the Department of Electrical Engineering and Computer Science at South Dakota State University is working on an important piece of the puzzle through a two-year grant of $261,503 from the U.S. National Science Foundation. He will develop algorithms to help these intelligent control systems use data efficiently and learn quickly. The long-term goal will be to integrate these algorithms into the navigation and scouting applications of a space robot such as the Mars Rover.
To do this, Ni, whose expertise is in computational intelligence, adaptive control, and machine learning, is collaborating with the NASA Ames Research Center in San Jose, Calif. He plans on incorporating the new algorithms on the NASA center's test bed in the summer. One doctoral student will work on the project.
In the long run, sophisticated intelligent learning control systems, such as the one developed for a space robot, could guide people to the nearest exits to evacuate a public building, such as a shopping mall, quickly in an emergency situation, such as a fire.
Electrical engineers working on the control system face two major challenges, data efficiency and learning speed. "I will develop a new learning and associate control framework," Ni says. This will involve "building the historic database and introducing a new sampling probability function to recall those useful experiences from the database."
For instance, in mathematics, a formula that a student learned in a previous class can then be applied to a current problem, Ni says. The student must recognize what is needed to solve the problem and then retrieve the formula, be it from a textbook or electronic source, and apply it to solve the problem. That's what a smart engineering learning system must be able to do.
"The space application is a very challenging task and I hope to contribute to some part of it," Ni says. In addition, Ni will integrate some of the research into his courses to give SDSU students insight into space robots and control technology. "Hopefully, this project can open up new career paths for our students," he says.
No entries found