Robots that Jump
Wednesday, May 14, 2003
Robot design: How come these robots look like boys?
In the future, advanced robots won’t be made to imitate human bodies – they’ll be works of art, or more specifically, part of interior decorating of homes. One can imagine looking through home and kitchen magazines at Home Depot circa 2020 and choosing a robot with the right style and color combination to go with, say “French Country.” This is not
design their living spaces to be comfortable and pleasing.
Two features of poorly designed robots seem to be: (1) Lots of wires hanging out, and (2) a male-shaped body. These metal boy fantasies won’t fly with interior decorators. Live with it – robots are going to quickly move away from hulking Transformer-like bodies to something more…elegant.
Fortunately, there is a designer in Japan named Tatsuya Matsui who appreciates this. Originally, he worked on the SIG Project, a humanoid robot that shows some of the best styling around. The design of SIG artistically captures and comments on humanoid form without trying to mimic people. The latter is a recipe for disaster – an almost-human robot will look like a movie monster. Related projects include PINO andMORPH.
Matsui’s robot style manifesto is at http://www.symbio.jst.go.jp/%7Etmatsui/. However, after leaving the SIG project he designed the POSY robot featured at Robodex 2002. This robot imitates a flower girl at a wedding, and is about as far from a robot monster as you can get. It has been used to sell perfume in Japan, which once again shows the difference in the US/Japanese mindset. The POSY robot is associated with SGI Japan. The home page shows a little Flash animation of the POSY robot talking to a bird – cute!
Matsui has also registered a website as the “International Robot Design Association” at http://www.iroda.org/ but at present there’s nothing there – might want to check it out in the future. There will be robot interior decorator catalogs someday, and this may be the start…
– posted by Pete @ 9:43 AM
Tuesday, May 13, 2003
Better artificial muscles will power robots that jump
In a report carried in AAAS Science, German researchers report that they have developed more efficient artificial muscles using metal alloys that change shape. These nickel/titanium strips move when voltage is applied. Previously, large voltages were needed to flex the metallic muscle, meaning that with repeated flexing large amounts of waste heat were generated.
The innovation reported uses platinum with tiny nano-sized holes. Because the platinum nanoparticles have a large surface area, they can store more static electric charge at lower voltages, greatly reducing waste heat. Weighing just one gram, the tiny muscles can lift 140 grams.
Such muscles are a shoo-in for advanced robotics. It is very difficult for standard electric motors to develop the torque necessary to handle walking, picking up packages, and more. In addition, electric motors are noisy and have more parts prone to wear and breakdown. New robotic muscles using the nanotechnology described above would be strong, efficient, and silent. Since they work the same way as biological muscles (meaning they contract during their work cycle) it will be easier to design robots by direct study of humans and animals.
This will cause a psychological effect as well. It is expected that use of metallic muscles will be widespread in toys, digital cameras, and the like. Use of metallic muscles will change the way these items move. Instead of whirring smooth movement, their parts will move the way an animal moves their limbs. The objects will appear more “alive” to humans and animals. There will be a change in the subjective “feel” of devices using this technology. In robots, the “stiffness” seen in walking will disappear. The ability to apply fast, sudden movement will make robots that replicate human dance moves practical.
– posted by Pete @ 8:16 AM
Monday, May 12, 2003
Q: Who wins robot wars? – A: how big is your company?
One of the best ways to understand why the US is likely to be last in consumer robotics is by looking at the size of companies involved. By definition, a robot that can jump – meaning it can easily navigate in two-leg environments versus wheelchair environments – is hugely expensive to develop to the feature and reliable operation needed for a consumer application beyond the “toy” level. At the very least, we’re talking about hundreds of millions of dollars.
This is something not appreciated by many looking at the robotic field. In their mind, someone will build a robot in their garage – a sort of R2D2 Apple II – and the revolution will launch. This will not happen. The robot revolution will not follow the personal computer revolution because the crucial components needed – robot bodies – cannot yet be mass-marketed.
The rise of PCs was allowed by the creation of integrated circuits which could act as a complete computer. A company like Apple Computer in 1978 just had to pick a dozen or so of these premade “chips” and wire them together to launch a revolution. Subsequent growth of the industry was driven by ever-faster chips, coupled with expanding memory and disk storage. There were very few “qualitative” changes – the PC revolution was comparable to someone revving an engine to ever-faster speeds. The heavy lifting of creating mass-market ICs had already been accomplished.
The reason this happened at all was because hundreds of millions, even billions of dollars had already been spent in the 1960s and 1970s to develop the integrated circuit to a cheap, mass-market product.
The mistake many roboticists (particularly those in ‘hobby’ robotics make) is that these very same chips will be the driver for the robotic industry. This is untrue. The key component of a robot, as opposed to a PC, is real-time processing of sensory data, coupled with automatic action based on sensory data via a robotic body. This does not simply require a chip – it requires a working body. There can be no compromise on the body, any more than a computer maker can use a half-realized IC chip.
And building a reliable robot body is a very expensive proposition. You can’t make up for it by making the robot more ‘clever.’ In fact, this is exactly the trap that so many “trash can on wheels” robots fall into. No matter how smart they are, they are still trash cans on wheels. The reason this is true is easy to understand. If I was paralyzed with no arms and only able to move in a wheelchair I’d make a dammed poor servant. No amount of cleverness could overcome this. And any robot ‘mind’ we can build in the near future will be considerably less intelligent. Without a reliable, flexible, adaptable, sensitive body it will have little real use beyond PC – like functions (getting email or consolidating remotes into one central “get me the DVD” system).
So, for PCs, a mass-market integrated circuit was essential, while for robots a mass-market capacity to create reliable, environmentally sensitive robot bodies is essential. The robot body stands to robotics like the integrated circuit stands to personal computing.
What would a robot body technology be like? Again, comparison to integrated circuits is helpful. While different IC chips have different functions, they are all designed in the same way and manufactured using a similar industrial process. Circuits are designed in a CAD/CAM program, and ICs are manufacutured using variations of the photolithography technique in use for decades.
A comparable “robot body” development system would include a commercial, “turnkey” CAD/CAM system for designing robot bodies according to physical laws, material properties of metal and plastic, weight, energy use, etc. The designer would be able to define potential wear surfaces, fix motor types, calculate where screws are needed, define how to make inner parts accessible to servicing, and more. The CAD/CAM system would also have to reliably simulate walking, jumping, getting up from a fall, etc.
Once design is done, a rapid prototyping system is needed to manufacture the candidate robot body. One can imagine the software including a system that rapidly compiles and organizes the product orders, vendors, and cost needed to build the body. Further out, a “3D” printer system could be used to assemble robot bodies on the fly. In either case, the cost of creating and testing the body would be comparable to burning a new chip design and testing it.
The reality is that today, neither of these components exist. Existing robot designers use CAD/CAM programs adapted from other purposes. The manufacturing system doesn’t exist. At present, building a robot body that can jump takes lots of money.
The “hundreds of millions” is still being spent. We are a long way away from the time when someone in a garage can design a robot complex enough to jump, buy cheap components, and build one on their own.
So at this stage, robot builders are in the position of chip builders at the end of the 1960s – making very complex and expensive devices, requiring large expenditures to advance their art. Big companies are needed. The success of a country’s robot initiatives will be dependent on the size of the companies doing the work.
On this bet the US loses hands-down. In Japan, industry giants like Mitsubishi Heavy Industries (Wakamaru) and Honda (Asimo) are spending the hundreds of millions needed. In the US, small companies are restricted to building very simple “trash can on wheels” bodies. Work on more complex components (arms with hands and fingers, for example) are scattered among US universities as individual projects, which frequently end after a few graduate students get their degree. The concerted application of money on a single project is missing.
Consider iRobots Roomba. The computer chips in this robo-vac are simple, and the behaviors of the robot are simples as well. It’s clear that most of the effort went into desiging the body, which is a major departure from a human-operated vacumn. Imagine if a legged device had been necessary. iRobot might have the tech know-how, but not the funds to build such a robot.
On this basis, it is pretty clear that Japan, and possibly a couple of other Asian countries are going to win the robot race – not because they’re smarter, but because they have large, “heavy industry” companies spending the money necessary to make practical robot bodies. Already, Sony has defined the standard “4-legged” robot body used in RoboCup. It seems likely that Honda and other Japanese companies will be selling humanoid bodies to US researchers before long. Anyone who doesn’t use their products will produce trash cans on wheels.
– posted by Pete @ 5:26 PM