The emergence of the robot brain

Author:  Corbaley Dean

Date:  February 2008

The first artificial mind is closer than we think: it is already here. Many researchers have built many different types of thinking machines, yet none so far have come to become the thinking, feeling, song-singing machines that we might expect. Nevertheless, scientists have given the next generation of robots new, remarkable features, the foremost being the ability to remember and to guess. Although these skills are present in even the most simplistic mammals, such abilities have made vast improvement to the performance of today's robots.

Most of us do not realize the extent to which robots are integrated in today's society, as robots are used in a wide array of functions. Bomb squads and SWAT teams use robots for defusing bombs and scouting, but these are all manually-operated and awkward. Recently, researchers in many laboratories around the world have developed a new type of software program that can allow a robot to navigate its own way. The term most researchers use to describe this new robotic ability is SLAM (simultaneous localization and mapping), and scientists are using it to create robots with a diversity of capabilities all over the world.



From the laboratory to the field, robots are excelling as they map, communicate, and navigate the environments laid out before them. Dr. Alfredo Weitzenfeld, a roboticist at the Autonomous Technology Institute of Mexico in Mexico City, created a robot rat (which is actually an AIBO Sony dog rewired). It can run a maze in a similar fashion to a real rat, remembering where it was before as it searches for its programmed reward'. But there are many more examples of cutting-edge robotic brains. For example, Durrant-Whyte of the University of Sydney and Phil Greenway of BAE Systems in Bristol, UK, built aerial drones that use video and radar images to build up a map of the terrain beneath. Four drones fly together in formation and exchange information between each other as they build their own maps of the terrain below. In another form of thinking and adapting robots, George Lee and colleagues at Purdue University, have created a robot that can navigate an office building, deciding whether or not to search certain areas determined on previous experience with similar looking areas.

"This approach makes sense," says Dr. Andrew Davison, SLAM researcher at Imperial College London, UK. "Rather than getting all the detailed information, they're learning static, repeating features of the environment." This allows robots to basically do what humans do naturally: to get the big picture' of a scene by collecting the important and recognizable features and making assumptions off of that. It is not a perfect system, but one that seems to work better than analyzing every detail about a scene.

To go a step further, researchers at Carnegie Mellon University (CMU) developed a robot that could navigate through unfamiliar terrain autonomously. The robot, named Groundhog, was equipped with a plethora of sensors to guide itself through an abandoned coal mine near New Eagle in southwestern Pennsylvania.

"Groundhog has to be ultra reliable because we don't have the option of taking control of it to correct its mistakes," said Robotics Institute Systems Scientist, Dr. Scott Thayer. Together with Professor William L. Red Whittaker, he teaches the Mobile Robot Development class at the Robotics Institute of CMU.

"The key is our state-of-the-art autonomous exploration and mapping software technology. The robot creates the map, makes its own plan, explores and comes back with useful information," says Thayer.

A robot that can navigate itself is a very versatile tool, and the technology and robotic capabilities are only going to increase from here. A robot that can recognize and adapt to its surroundings would be useful in a multitude of situations from urban disasters to battlefield applications, and someday even planetary exploration. Some scientists have hopes of a much smaller robot that could map and navigate the human circulatory system, allowing a whole new form of angiograms and angioplasty, the charting and repair of our blood vessels. If a robot could learn the routes of a particular human's circulatory system, it could travel them and report on any clots, scars, or other abnormalities that it may find. This could help doctors diagnose and possibly treat circulatory problems much more efficiently than they currently can.

Robots have not come anywhere near the intelligence of a sentient creature, but they have come a long way from their humble beginnings. Now robots' processing capabilities can rival that of simple mammals, and research shows no signs of slowing down. The way things are going, robots will be singing, dancing, and waxing philosophy in no time.

Written by Dean Corbaley

Reviewed by Nira Datta

Published by Pooja Ghatalia.