Emerald Group Publishing Limited
Copyright © 2000, MCB UP Limited
May the Force be with you
May the Force be with you
Welcome to the new millennium! I remember once being told that the best thing about company accounts is that they are done annually, so no matter what disasters have befallen the previous year, once you get to the year end you can put it all behind you and face the new year with a spring in your step. If you can do that at the end of a year, then the opportunities at the end of a millennium must be even more extensive.
Given a clean sheet of paper, would you design your next robot by tinkering with its predecessor or instigate dramatic and fundamental changes? If the latter course appeals, then I would expect force feedback to feature quite highly on your new list of design requirements.
To those fortunate enough to have never needed to consider force control, the problem may appear trivial, and I can only recommend that they read our tutorial "Force control basics", by Dr Edwin A. Erlbacher of PushCorp.
There are basically two fundamental problems with force control. The first is knowing the force that you are applying, which should be relatively easy, but is greatly complicated by gravity. And the second is being sure that you actually want to apply any force at all.
To illustrate the problem, let us take the example of a bit of flash on the edge of a casting. The aim is to remove the flash without unnecessarily removing material from the rest of the casting. The desired profile is known, but the casting may not be positioned too precisely, and also the size and location of the flash are unknown. A pre-programmed path around the profile with a milling tool will soon result in a broken tool and damaged casting. The edge needs to be sensed and tracked so that the flash is accurately located and then removed using force feedback to ensure that the tool is not broken. It sounds straightforward enough, but finding a good, robust solution is far from trivial, as our paper, "Robotic fettling and automated path acquisition", by B. Shirinzadeh, P. Teoh and M. Roberts indicates.
The real fundamental problem is that our robots are not equipped with sufficient sensors. If they were able to see and feel and monitor internal stresses, temperature, vibrations and sounds, then our flash removal problem, and many others such as robot calibration, product location and collision avoidance would become much simpler.
It is quite possible today to provide a robot with all of the senses outlined in the previous paragraph. And indeed, it should be quite easy, except that such additions are effectively external "add-ons". Mention them to a robot manufacturer and you are likely to be told: "We wish people would stop asking us for this - we keep telling them there is no demand!"
My first PC was an Apple II, and in those days of garden shed entrepreneurs, it was possible to buy add-in cards that enabled it to measure external temperatures, control train sets, recognise spoken commands and act as a half-decent music synthesiser. If you saved up enough, you could also splash out on a 5MB external hard disk drive. Although this lot gave hours of fun, the tangle of wires and plug-in cards did not add to the system's reliability and in the all too brief instances of "up time", these utilities performed individually, one at a time and with no chance of integration.
My latest PC is comparatively a marvel of integration. I can talk to it and, via a built-in modem, to the rest of the world. It talks back, plays music CDs and I can (should I get the urge) dabble on the stock market while writing an editorial and control a robot halfway around the world. This has been made possible and, more importantly, easy by the integration of these various systems within a standardised computer operating system.
Industrial robots have been around for quite a bit longer than PCs and yet most have no more senses than the first spot welding Unimates. My wish for our next generation of robots is that manufacturers adopt a standard "open" operating system and include an array of fully-integrated sensor capabilities. If they can do this, then I am sure that new markets and opportunities will quickly develop.