Simple progress
Article Type: Editorial From: Industrial Robot: An International Journal, Volume 36, Issue 5
I have been giving some thought recently to the reliability of computer software. I am quite highly qualified in electronic engineering. I have spent many more hours than I care to remember writing software programs for all sorts of weird and wonderful equipment. But when it comes to using my Windows XP-based desktop computer, I am frequently frustrated by it crashing for no good reason at all, and the helpful messages that get displayed only add insult to injury.
It is not just me either – my friends all report the same anguish. A few days ago, my new Netbook started “donging” for no reason and I have spent literally over a day on the phone to the manufacturer and I have even had a technician around who replaced the motherboard and hard drive and re-installed the system software – and still it keeps “donging.”
If you compare this level of reliability and user friendliness with today's cars, which are arguably just as sophisticated, then the car comes out well ahead. I do not know the statistics, but I would expect that there are roughly the same number of desktop computers in the world as there are cars, and each has a roughly equivalent life expectancy.
So, why is it that cars are so much more reliable? It has not always been thus – I spent a good proportion of my student years underneath my Fiat 127 trying to bleed air out of the braking system, or tuning the engine by adjusting the ignition advance and retard with a screwdriver and a spanner or two.
Where would we be if our cars were as unreliable as our desktop computers?These days I expect my car to work, and what is more I can hire a different one and instantly drive it away. This is user friendliness taken to an advanced degree. I no longer need to know how to fix it because it does not go wrong– I can just use it.
I suppose that in its defense it could be argued that PCs are unreliable because we ask them to do so many different things.
Around the same time that I was getting smothered in brake fluid the mean time before failure of a Unimate robot was a few hundred hours. These days the MTBF of industrial robots is many tens of thousands of hours – or once every few years. This very significant improvement has come about because it had to. Any manufacturer still supplying original Unimate levels of reliability would have ceased trading years ago.
Our main theme for this issue is Humanoid robots and I would say that it is currently unreasonable for us to expect these new developments to have anything like the same level of reliability as their modern industrial counterparts. However, we do need them to be as reliable as this if they are really going to make a big impact on our lives. The early Unimates were shielded with high metal fences to keep people well clear of them and safe – and the necessity for this was first brought to public attention by the tragic death of Robert Williams, killed in 1979 at a Ford plant in Michigan, USA.
How can we ensure that humanoid robots are both highly reliable and safe to work with?
It is often said that we buy features but like simplicity. And this delusion has led us blindly down the road to the desktop PCs that cause us so much anguish. I hope that engineers working on tomorrow's humanoid robots will concentrate on making robots that can do simple tasks reliably rather than everything not well at all.
Clive Loughlin
