Talk:Advanced automation/Self-maintenance and repair

I'm interested to know if there is much evidence of this on the horizon. The ideas in this article make sense to me, but is there any evidence that they are starting to be put into practice? I know there is autonomic computing, which is essentially self-maintenance and repair for software, but are there existing examples of self-maintaining hardware? -- Balatro

Good question. Not much that I can think of, yet (must have another trawl at some point, last one was a few years ago). This is simply my logical conclusion as to what seems technically possible. It is however a pretty key technology I think for an advanced post-scarcity society. I am keen to demonstrate that this is possible actually by having a working machine of some sort where a failure can be simulated on command and a repair system is able to swap out any of the components - a good AdCiv pilot project (had better get some funding)!

Why hasn't it happened yet? Well a few reasons I think, and some are a bit chicken-and-egg (in the commercial world anyway). Firstly machines at the moment aren't designed to be repaired by robot, as plenty of dexterous humans around. Initially designing self-repairing machines would need a little more human effort in the design process (however this aspect of the design could be automated too). Secondly the repair robots simply don't exist yet either. PR2 might be a good template platform, although PR2 is not actually open-source - only ROS, which it's running. It is just a research platform currently and not optimised for repair tasks. Currently robots are too expensive and a too primitive still, but this is changing fast. Also there would probably need to be standards in place for commercial machine manufacturers / operators and repair-bot builders. Although this should not be such an issue in the world of open-source.

We are also only just getting to the technological level where we can have tiny sensors embedded in all components giving information at the resolution required. Managing the huge data flow within a complex system would be no mean feat either. All possible with current know-how though as far as I can see, it is just very early days. There are very advanced diagnostics and control systems already in places like power-plants, and fairly complex and accurate automated assembly in factories (see video of Dyson's automated motor manufacturing - link bottom-left of first section). It is a matter of putting the two together with accurate 3D localisation technology for the repair bot. -- Charles.

General Electrics and others car manufacturers are building systems that can diagnose engine problems and communicate them wirelessly to service professionals, or even order parts. http://www.youtube.com/watch?v=3YUjlKA6tVU#t=49m17s --Balatro 06:39, 21 April 2011 (CEST)

Whatever about automatic replacement of parts, self-healing materials are a big research area, and will almost certainly be in machines in the next 10-20 years. Like self-replication, healing is a characteristically biological trait that technology is now getting. http://www.kurzweilai.net/self-healing-autonomous-material-mimics-bio-systems --Balatro 14:14, 10 December 2010 (CET)

Yes self-healing is certainly interesting. For the near-term future I think it limited to simply creating more robust materials rather than having sophisticated healing properties needed for failures in more complex systems with moving parts. My main point here is to try and illustrate that self-maintaining and auto-scaling systems, which will be hugely powerful in terms of circumventing current scarcity-based economics, is possible with technology that we have today, and we just need to get on with it and develop open-standards for these kinds of machine systems. Sure, self-healing will be important at the material individual component level though. All the of the self-healing materials I've come across so far are interesting, but are actually weaker than their traditional non-self-healing counterparts (that is until there's a failure!). CharlesC 19:23, 17 December 2010 (CET)