Technical problems need to be overcome before we can buy a robot butler, but technology can surprise us with breakthroughs
BY RICHARD F. ZARR
National Semiconductor
Santa Clara, CA
http://www.nsc.com
We have heard the hype: “Robots in the home will change life forever, providing every convenience and comfort.” Okay, so where are they? Major hurdles must be overcome in various technologies before the robot butler will be in your home. These areas include sensors, motors, and power sources as well as software and artificial intelligence. Let’s examine a few areas and see how research is moving closer to the household robot.
Sensors
In the world of industrial control, sensors can be found everywhere. They are usually specially designed devices targeted for applications such as temperature, pressure, pH, stress, and even more complex functions such as imaging. If we examine how a human interacts with an object, it becomes quickly apparent that we employ multiple senses to understand and manipulate it. Just watch a baby playing with a block of wood. Now look at another example on the other end of the spectrum: gymnastics. Gymnasts use amazing strength and extremely fine motor and sensory skills to leap and twirl their bodies precisely through complex routines. If anything changes their perception of the environment, the results can be catastrophic. This too is true of a robot.
Accurately sensing the environment is essential for a general-purpose robot. A human’s skin contains all kinds of sensors, including hair follicles that respond to pressure, temperature, and contact with other objects. To reproduce this network would take millions of sensors all feeding a continuous stream of data to our processing engine. Today, a lot of research is going on to develop a skinlike material that can replicate our sense of touch. Without the sensitivity and variation of mechanical stimuli, a robot would effectively have a poor sense of touch: a show stopper for our home robot. An artificial skin that can repair itself, sense temperature and pressure, and is warm to the touch, flexible, and resilient would be a major breakthrough for both robotics and prosthetic limbs.
A company called Peratech (www.peratech.com) has pioneered a new material based on a patented technology called Quantum Tunneling Composite (QTC). The material is made of spiky conductive nanoparticles suspended in a binder that isolates them from one another. The quantum view of electrons allows them to tunnel across barriers if the wave property of the electron does not decay to zero before exiting. Using this principle the QTC material can change the barrier potential through applied pressure affectively changing the materials resistance. QTC material is commercially available and is a very good candidate for creating a pressure-sensitive skin.
Other researchers in Japan are combining (or fusing) sensor arrays to provide not only a sense of touch, but of temperature as well. With the ability to fuse sensors, robot skins could detect light or even sound — something human skin cannot do. Touch is only one sense humans possess; there’s also an amazing sense of sight, hearing, smell, and taste. Other animals, such as pigeons, can sense the earth’s magnet field and use it for finding direction. A practical general-purpose humanoid robot could also include sensors for toxic gases, biological agents, radiation, and other phenomenon beyond the normal senses of humans.
Motors and actuators
Even if the sensing problem is overcome, the issue with motion remains. Motion in humans is accomplished through the application of force across a joint. This can also be seen in machines that use hydraulic cylinders such as a backhoe. Electric motors, in linear and rotating versions, have been used in many robotic applications, but they suffer from problems such as weight, size, and mechanical conversion efficiency.
In a humanoid robot, rotating motors can be useful. However, a better solution might be a muscle motor — or a material that contracts with applied electric field. Humans have roughly 640 muscles that operate in pairs. Our muscles change their length roughly 20% to 40% when activated and contract at a rate of roughly 20% per second. This is pretty good performance, which can be verified while watching most athletic or dance competitions.
However, researchers at the University of Texas at Dallas (UT Dallas) have discovered a synthetic muscle material made from carbon nanotube aerogel sheets (Science Magazine, March 20, 2009). These artificial muscles (see Fig. 1 ) can change their length 220% and contract at a rate of 30,000% per second! They also have a tensile strength greater than steel and can operate in extreme temperatures (–196° to 1,538°C), allowing robots using this material to operate in the harshest environments. Once this technology is available to robot builders, robots and prosthetic limbs will have newfound flexibility and strength.
Fig. 1. How artificial carbon nanotube muscles work.
Brains
As humans we take for granted the difficult process of simply walking across a street without getting hit by a car. The amount of data our brains process every second is overwhelming, but most of the processing is automatic, without conscious effort.
For machines to mimic this behavior, many problems need to be solved. Mostly our immense knowledge of the world and the way we access it is still far beyond what machines can do. World knowledge database projects such as Cyc, WorldNet, and MindPixel are just a few examples of attempts to wrap our common knowledge — that which we learn over a lifetime — into a database that computers and robots could use to navigate our world.
There are physical limits to our current transistor technology due to the physical nature of the materials used and the architecture of the structures. Next-generation digital processes will use quantum-tunneling devices or other quantum-effect devices to pack fantastic amounts of circuitry into a single device (see my blog post, “The World Electric — Part III” at www.energyzarr.com for more detail). The added benefit will be a vast reduction in power consumption — something greatly needed for portable devices and mobile robots.
Beyond the hardware needed to store and access a vast knowledge base is the ability to use that data to reason and to interpret language. Humans learn this as they develop, and it is incredibly difficult to mimic. Imagine what a robot butler would need to know to understand the phase “Find me a casual restaurant that suits my taste.” The words “find” and “suits” are easy to interpret, but the phrase “casual restaurant” implies a meaning that humans understand — here referring to the dress attire being casual. It seems so simple, but we do this hundreds of times every day without flinching. Computers have much to learn if they will interact with humans as we interact with each other.
Power sources
Last, but not least, is the problem of powering our humanoid robot. Machines that work and coexist with humans must be capable of continuous operation for many hours without running out of energy. Current battery technologies are simply not advanced enough for this kind of operation unless the total power consumed by such machines is greatly reduced. Today, electric motors and the computers that run them can potentially use hundreds of watts of power, quickly draining the batteries.
Electric vehicle research may actually help this area since batteries and their ability to store energy is central to a practical EV design.Research is ongoing in several areas, including supercapacitors that use carbon nanotubes to increase the charge capacity simply by increasing surface area (see Fig. 2 ). Other technologies are focused on improving charge time of lithium-ion batteries to minutes or even seconds instead of hours, so our robot might need charging breaks a few times a day for 5 minutes.
Fig. 2. Comparison of conventional capacitor and carbon nano-tube super-capacitor.
Among all of these new technologies, many technical problems need to be overcome before we can purchase a robot butler at our local technology store. Technology has a way of surprising us, and breakthroughs — perhaps one in artificial intelligence — will catapult our home humanoid robot to the production line faster than expected. ■
Learn more about National Semiconductor