At the forefront of Artificial Intelligence

An Introduction to Robotics

Most of Artificial Intelligence will eventually lead to robotics. Most neural networking, natural language processing, image recognition, speech recognition/synthesis research aims at eventually encorporating their technology into the epitome of robotics - the creation of a fully humanoid robot.

The field of robotics has been around nearly as long as Artificial Intelligence - but the field has made little progress. This is only natural, since the field not only attempts to conquer intelligence, but also the body that embodies it - a formidable task indeed! Robotics, though, is not just about humanoid robots; but also about their commerical applications in manufacturing, safety and hundreds of other fields. Let us back-track though, and look at what could consistute a robot?

What is a Robot?

According to the Oxford Dictionary, a robot is an "apparently human automaton, intelligent and obedient but impersonal machine". Indeed, the word robot comes from robota, Czech for 'forced labour'. Yet, as robotics advances this definition is rapidly becoming old. Basically, a robot is a machine designed to do a human job (excluding research robots) that is either tedious, slow or hazardous. It is only relatively recently that robots have started to employ a degree of Artificial Intelligence in their work - many robots required human operators, or precise guidance throughout their missions. Slowly, robots are becoming more and more autonomous.

The difference between robots and machinery is the presence of autonomy, flexibility and precision. Indeed, many few robots are mere extensions of machinery - but as the field advances more and more, the current 'fine line' will widen more and more. To understand more of what robots can include, let us look at some examples.

Current Projects

There are many interesting robot projects. RoboMonkey is a great example - a robot built to emulate the gibbon. This incredibly agile robot can swing from bar to bar (fixed distance) by using its body to give it the correct momentum. The robot learns from its mistakes, and will adapt accordingly. But for me, the most interesting projects are being conducted at MIT's robotics laboratory. Two major robot projects are in progress (and have been so for many years) - Cog and Kismet.

Cog

When MIT started the Cog project, their primary ideology was that the robot must be built from the bottom-up. The robot would be taught all the necessary data it needed - with very little explicit programming. Cog will eventually be a complete humanoid robot; as it currently stands (sits) it has a head, torso and arms all with proportional degrees of freedom.

As of yet, Cog does not perform any higher-level functions of a human, nevertheless the current robot is quite incredible. Cog can track target smoothly with his head (smooth-pursuit), he can move his head if a stimulus is provided somewhere in his field of vision (saccade to motion). Cog even has a vestibulo-ocular reflex (the ability to keep the eyes in a fixed position when the head moves). Cog can recognize faces and can detect whether eye contact has been established by using his high-resolution cameras. The robot has also learnt how to reach for targets through trial-and-error and can even imitate simply facial movements.

The philosophy behind Cog is essential to its success, and continued success. The following passage really demonstrates how the Cog team feels:

"…We believe that classical and neo-classical AI make a fundamental error: both approaches make the mistake of assuming that because a description of reasoning/behavior/learning is possible at some level, then that description must be made explicit and internal to any system that carries out the reasoning/behavior/learning. This introspective confusion between surface observations and deep structure has led AI away from its original goals of building complex, versatile, intelligent systems and toward the construction of systems capable of performing only within limited problem domains and in extremely constrained environmental conditions.…We believe that [our] abilities are a direct result of four intertwining key human attributes: developmental organization, social interaction, embodiment and physical coupling, and multimodal integration of the system…"

Kismet

Kismet is aimed at helping research robots and social interaction with human beings. Kismet is only a head, but much more detailed than its Cog counterpart. Kismet's head consist of two eye's (with embedded cameras), ears, a mouth, and eyebrows. These can be combined to produce numerous facial expressions denoting Kismet's feelings and emotions. Kismet can be stimulating using toys such as a slinky - over stimulation will upset him, under-stimulation will bore him, but just the right amount will make him happy.

The Cog team believes that social interaction is key to helping robots learn and grow - just like babies. Therefore, Kismet is a precursor to teaching Cog all about life! Kismet has a dedicated 'trainer' that plays and interacts with Kismet, and watches his output (Kismet is controlled by a Pentium processor) on the monitor to gain a further understanding of the robots interal states.

Robots at Home

Robotics is slowly making its way into the home - either through leisure, or actual commercial home-based bots. Recently, Probotics released the world's firt true personal robot - Cye. Cye allows its human operator to create a map of the environment (using a Windows interface) and download it via an IR link to the robot. The robot will then be able to navigate the area doing various tasks - including vacuuming! Consumer robots, though, have not yet made a big impact. So-called leisure-robots are.

Recently, Tiger released a furry little toy called a Furby. These toys were touted to contain some impressive artificial intelligence, plus the ability to communicate, talk and learn from other Furbies. Furbies have an array of sensors - light sensors, tilt sensors, a microphone, various strategically-placed buttons, and an IR communication device. Furbies have their own language, but can apparently learn any other language. Once they have learnt the language, they will start to teach it to other Furbies in their vicinity. Furbies are extremely well-priced now, and a robotics version The Gigabot is now available. Sony released their own version - a much hyped, expensive electronic dog, Aibo. The dog does various tricks, has a proximity sensor to avoid obstacles, and can even pick itself up when it falls over! The unit was released in several short bursts, with Sony releasing about 10,000 each valued at about $2,500 each!

Below are three pictures of (left to right) the Cye SR, a Furby, and Aibo.

Conclusion

Robotics is an absolutely fascinating field that interests most people - AI buff or not. As research from more serious robotics projects such as Cog and Kismet filter down into the commerical arena we should look forward to some very interesting (and cheap) virtual pets like Aibo and the furbies. Hopefully, commericial home-based robots will also be avaible for a price not more than an expensive vacuum cleaner. With computers becoming more and more powerful, interfacing home robots with your computer will become a reality, and house work will (hopefully!) disappear.

Last Updated: 10/03/2000

Article Toolbar

BibTeX entry

Glossary Terms

- Humanoid
- Neural Network
- Natural Language Processing
- Speech Recognition