Read our COVID-19 research and news.

Fuzzy Logic, Adventures in Artificial Intelligence

Will robots ever be able to learn the way humans do? After all, gathering data about one's environment is the easy part; the difficult part is being able to evaluate that information and adjust one's response to it. Answering the call to address this highly complicated and technical question is 31-year-old Ayanna Howard (pictured left), senior member of the Telerobotics Research and Applications Group at NASA's Jet Propulsion Laboratory (JPL) in Pasadena, California. (JPL is governed and administered by the California Institute of Technology for NASA.) She is developing a software program system that emulates human behavior for use in a Mars robot rover. The robot will search the surface of the Red Planet for evidence of water and life and will pave the way for human exploration.

Howard has always been driven by her desire to "build seamless human-robot interface systems." "As a third grader I watched The Bionic Woman and became interested in robotics and making robots smarter," she says. Her parents took the hint and gave her an Erector set and an Easy-Bake toy oven for an especially memorable Christmas. Following a cake recipe or the instructions for building a structure sparked her curiosity about logic and made her appreciate hands-on work. She also learned a truth that would continue throughout her career, "Knowledge and success come through experimentation."

As an undergraduate at Brown University, Howard worked in a robotics lab studying locomotion and engineering, but her internship at JPL in 1990 introduced her to the world of artificial intelligence (AI) research. She enjoyed the experience so much that she returned to JPL every summer. After completing her B.S. in computer engineering at Brown, Howard decided to turn her summer research sessions in AI into graduate-level study. She worked 30 hours a week while earning her doctorate in electrical engineering as a full-time Graduate Engineering for Minorities (GEM) Fellow at the University of Southern California in Los Angeles. "It was difficult to say where school ended and JPL began. I was doing 'hard-core' research with neural networks, machine vision, and health monitoring for faults in satellites beginning at age 22, while classmates were only doing 'school' research," she says.

Howard's educational background allows her to understand the level of sophistication needed to train robots. She admits that some suboptimal solutions may leave AI-based robotics incapable of operating in difficult (e.g., realistic) conditions, such as a rover that can only navigate smooth terrain. "I want the robot to react like a student: being able to do certain things by itself but still needing teacher input for other tasks. Ultimately, we want robots to be an extension of ourselves," Howard says. Concepts such as fuzzy logic, neural networks, and genetic algorithms are used to introduce what she calls "humanized intelligence" (see box).

Humanized Intelligence

Fuzzy Logic

Ayanna Howard defines fuzzy logic as extrapolating how humans think and learn with linguistic terms and conditional statements and then mapping the human information to the robot, machine, or application. A good example is how children perform technically difficult tasks, such as walking. "Kids can do amazing things, yet they haven't taken calculus. They use inaccurate measurements and imprecise knowledge to function effectively. In effect, they are using fuzzy logic," Howard says. The fuzzy logic model that Howard uses for rover navigation looks for free paths in the terrain, unlike the classic model that extracts individual elements (e.g., rocks, cliffs). Thus, the robot identifies the safest terrain from the number of free paths and follows it.

Neural Networks

A complement to fuzzy logic programming is neural network technology. A neural network is a series of interconnected processing elements (like neurons) based on the mammalian nervous system. It emulates properties of the nervous system, including adaptive biological learning, with mathematical models (Pacific Northwest National Laboratory). The rover neural network bases its decisions on values called strengths or weights that are used to learn human thought patterns through training sessions. More information on this topic may be found at " What Is an Artificial Neural Network?."

Genetic Algorithms

Neural network training sessions can also use genetic algorithms for adaptive learning. Genetic algorithms use the evolutionary principles of natural selection, crossover, and mutation to select the most favorable solutions. See NNUGA: Neural Network Using Genetic Algorithms by Omri Weisman and Ziv Pollack.

Howard's success has been well documented, including being selected as one of the top 100 innovators of the year in 2003 by MIT's Enterprise Technology Magazine and receiving the Lew Allen Award for Excellence in Research, the highest honor at JPL, for her work with AI and machine vision in 2001. She has published over 40 journal articles, conference papers, and book chapters. She also received a grant allowing her to develop the AI Toolkit, an educational software package designed to teach her student interns about fuzzy logic, neural networks, and genetic algorithms. The AI Toolkit shortens the otherwise 2-week learning curve for the interns with simplified examples of the robotic rover. The software is now available to the general public.

Howard credits her phenomenal success at such a young age to her family, especially her husband, for providing crucial support. According to Howard, "I've known people who've tried to do [what I've done] but didn't make it." Being the only person of color or one of a few made things difficult at times. Her success was not easy. "I was sometimes asked if I was in the right place or was told before asking a question that the secretaries were meeting down the hall, but I realized I was an excellent researcher and that my work spoke for itself," Howard says.

Although her work keeps her busy, Howard is committed to imbuing the same positive attitude about math and science to the girls in the Pasadena, California, community where she grew up and attended public school. She serves as chair and founder of the Pasadena Delta Academy, a subsidiary of the Dr. Betty Shabazz Delta Academy, a national nonprofit initiative founded by Delta Sigma Theta Sorority Inc. to educate "at-risk" girls in science, math, and technology. She says, "Balancing work and personal life is difficult and requires a personal sacrifice. You have to make a conscious effort to schedule your time for what's important."

Clinton Parks is a contributing writer for MiSciNet and may be reached at