Named Diego-San, the robot is used to conduct research on the brain development of children. It is part of the Machine Perception Laboratory — part of UCSD’s division of Calit2 and the Institute for Neural Computation at UCSD — whose mission is to create a robot that can learn and develop sensory-motor and communicative skills of infants.

Dr. Javier Movellan, principal investigator and founder of MPL, created Diego-San in order to study how babies acquire problem-solving skills that are, traditionally, difficult to reproduce with robots.

“The idea came from the fact that many problems in robotics and artificial intelligence have been very difficult to solve with classic A.I. approaches,” Movellan said. “Yet babies solve, within one year of life, the most complicated problems effortlessly.”

Movellan said he created the project to understand the complex brain development of a baby when it learns the motor and social skills necessary to survive.

The four-foot, 66-pound humanoid robot is equipped with a head that houses 20 moving parts that make facial expressions without speaking, two high-definition cameras for eyes, an audio speaker for a mouth and a six-axis accelerometer for ears, to detect orientation and movement.

The robot also has five-fingered hands that can hold objects like water bottles. It has sensors that detect the pressure points placed on different joints in its body, and can stand up from a chair.

The main goal, according to Movellan, is for the robot to carry out all these functions independently by integrating motor control functions in social interactions with only the help of its sensors and activators.

MPL is currently in its second phase of the project — which includes developing the control electronics and software — and is trying to replicate the way babies learn motor movements.

Specifically, the researchers want the robot to be able to recognize facial expressions, learn and recognize the appearance of its caregivers, reach objects on its own without human aid and use gestures like pointing, reaching and grasping for items.

To do so, researchers are currently working on an algorithm that integrates a multidisciplinary approach of bringing together the fields of physics, electronics and psychology.

“The robot really is an instrument for research and our goal is to push the limits of robot control,” Movellan said. We want it to have similar properties to that of a human being and the level of complexity similar to that of a human being.”

The robot has already been publicized in a newsletter published by Kokoro, a company in the Sanrio Group that specializes in animated robots and automated machinery. It has also been featured in several tech blogs, such as www.gizmodo.com, to mostly negative responses on the baby’s appearance.

Of the responses from blogs, Movellan said, “One of my students told me about this. I was surprised that this made it through the Internet.”

He said did not expect the newsletter to reach the Internet. But now, he said he sees the feedback as an opportunity to improve the robot’s design. For example, he plans to change the aesthetics of the face by working with Hanson Robotics, who collaborated with MPL on the Einstein robot.

Movellan responded to some of the blogs, clarifying that researchers were still in the initial phase of developing the robot’s body (which was constructed by Kokoro, along with the head).

After the developmental phase, MPL is planning to train the robot.

“It’s been great [because it’s] such a challenging project that tests you in so many different ways,” Movellan said. “What I expect from the project is a better understanding of the nature of human intelligence.”

The research began in September 2008. Diego-San is part of Project One, funded by the National Science Foundation.

As far as a projected date for completion, Movellan said that — though the funding will be cut off in 2013 — work on Diego-San is likely to continue much longer.

“The project is part of an ongoing effort to understand the development of intelligence in infants from a computational point of view,” Movellan said. “The effort is likely to continue for as long as I am alive.”

Readers can contact Regina Ip at [email protected].