Self-Healing Circuits for Deep Space

by Paul Gilster on March 18, 2013

Computer failures can happen any time, but it’s been so long since I’ve had a hard disk failure that I rarely worry about such problems. Part of my relaxed stance has to do with backups, which I always keep in triplicate, so when I discovered Friday afternoon that one of my hard disks had failed — quickly and catastrophically — it was more of a nuisance than anything else. It meant taking out the old disk, going out to buy a new one and installing same, and then loading an operating system on it. Because I do 90 percent of my work in Linux, I opted for Linux Mint as a change of pace from Ubuntu, making it the tenth version of Linux I’ve used over the years.

My weekend was mildly affected, but the new disk went in swiftly and the operating system load went without incident, so I was still able to get to two concerts, one of them an absolutely brilliant handling of Elgar’s ‘Enigma Variations,’ and to see the new Tommy Lee Jones movie ‘Emperor.’ Hardware failures in the midst of an urban environment, and with adequate backups on hand, are thus easily handled. But then I started thinking about robotics and deep space. Ponder the hardware failures that are inevitable on missions lasting decades or even centuries. An unexpected failure in a key circuit could wreck a lot more than a weekend on such a probe.

From Wardens to Self-Healing

Remember the ‘wardens’ that were built into the Project Daedalus plan? They were designed to take care of the vessel on its 50-year run to Barnard’s Star, an acknowledgment of what happens to complex systems over time. These days we’re focusing in on self-healing electronics that can repair themselves in microseconds, integrated chips that spring back from potential disaster, rebuilding themselves faster than any human intervention could manage. Members of the High-Speed Integrated Circuits laboratory at Caltech have been experimenting with self-healing integrated chips that can recover all but instantaneously from serious levels of damage.

Hajimiri-SEM-009-NEWS-WEB

Image: Some of the damage Caltech engineers intentionally inflicted on their self-healing power amplifier using a high-power laser. The chip was able to recover from complete transistor destruction. This image was captured with a scanning electron microscope. Credit: Caltech.

The chips in question are high-frequency power amplifiers useful for communications, imaging, sensing and other applications. Each of these chips holds more than 100,000 transistors along with a custom-made application-specific integrated-circuit (ASIC) that monitors the amplifier’s performance and adjusts the system’s actuators when changes are called for. The idea is to let the system itself determine the need to use the actuators without humans overseeing the process. Researchers therefore target the chips with a high-power laser over and over again, observing the chips as they come up with split-second workarounds to the damage.

“It was incredible the first time the system kicked in and healed itself. It felt like we were witnessing the next step in the evolution of integrated circuits,” says Ali Hajimiri (Caltech). “We had literally just blasted half the amplifier and vaporized many of its components, such as transistors, and it was able to recover to nearly its ideal performance.”

This Caltech news release compares the healing properties of these integrated-circuit chips to the human immune system, which can likewise respond quickly to a wide range of attacks. Interestingly, the team discovered that the amplifiers with self-healing capacity consumed about half the amount of power as standard amplifiers, while their overall performance was more predictable. By demonstrating self-healing in a highly complex system like this one, the Caltech researchers have shown that it can be extended to many other electronic systems.

All this is good news for our starship. We naturally think about catastrophic problems that damage parts of the circuits, but when we’re thinking long-term, the issues are likely to be more subtle. Problems can emerge as continual load stresses the system and causes changes to its internal properties, while variations in temperature and supply voltage can also degrade operations. For that matter, variation across components can play a role, making an electronic system with a built-in immune function an insurance policy for deep space robotic missions.

Meanwhile, my own computer operations continue with extensive human intervention, though I’m pleased to see that the new hard disk I installed checks out perfectly. We are all learning through experience how our lives are supplemented and changed by digital technologies. But robotic probes operating at the edge of the Solar System and beyond have no repair team on staff to open up a housing and plug in a new chip, We’re now learning that beyond redundancy and backups a new set of tools are emerging that will keep long-haul missions healthy.

The paper is Bowers et al., “Integrated Self-Healing for mm-Wave Power Amplifiers,” IEEE Transactions on Microwave Theory and Techniques Vol. 61, Issue 3 (2013), pp. 1301-1315 (abstract). Thanks to Eric Davis for the pointer to this work.

tzf_img_post

{ 15 comments }

James D. Stilwell March 18, 2013 at 12:06

Wishing Paul Gilster would revisit Jack Williamson’s The Humanoids….
AI still has quite a way to go to catch up with Jack’s 1948 imagination but solid progress is happening on so many different fronts as to make the humanoids a possibility in our time….AI is making faster progress than Jack
would like….But like it or not here’s the domain of the Fourth Industrial Revolution slipping into our revolutionary century….JDS

Peter C. Chapin March 18, 2013 at 12:42

I think this is fascinating work and every bit as essential as research into advanced propulsion systems. Building mechanical and electronic devices that can function without human intervention for decades or centuries is every bit as challenging as some of the other technical barriers facing interstellar travel.

Constructing correspondingly reliable software is also a major challenge that shouldn’t be underestimated. Formal methods are being used today to boost the reliability of software above that obtainable using traditional development methods. However, new technologies in software development and verification (self healing software?) will likely be needed before we can get serious about mounting any long duration space mission.

Alex Tolley March 18, 2013 at 15:37

Machines that have self repair properties like living systems would be useful not just on spacecraft. But as Joy noted earlier, the interesting thing is how the spacecraft is developed as a living organism. This solves many problems, including having lots of spare parts, trained people to repair/fix/change the parts, etc. We should be thinking in these terms for our large buildings and possible arcologies on earth (and eventually elsewhere in the solar system).

RonSmith March 18, 2013 at 16:32

Perhaps this is a quible, but I dislike the use of the phrase ‘Self-Healing’ for this quite remarkable technology.
If I cut myself my skin eventually heals. If I loose my eyesight in an accident, my brain learns to compensate with my other senses, as best it can.
What is being done in this technology more closely resembles the latter than the former.
To me true ‘Self-Healing’ would be when a circuit can, somehow, stitch the remaining parts of itself together and regrow the lost parts.
This technology is a major step forward, but I just don’t see ‘Self-Healing’ as an acurate name for it.
(Stop being so pendantic Ron!)

Paul Gilster March 18, 2013 at 16:53

James D. Stilwell writes:

Wishing Paul Gilster would revisit Jack Williamson’s The Humanoids….

Agreed. Won’t be long…

nat March 18, 2013 at 16:59

The current situation for the Curiosity rover on Mars might be an interesting case in point. It’s been down for around 2 weeks now on that distant world, due to a problem that is widely suspected from a cosmic ray blast in one of its memory banks. But the rover’s computer is set up to jump into “safe” mode when a severe glitch happens such as this, then waits for instructions from the distant humans. The diagnostics and corrections the humans are doing now on Earth maybe could be converted all into code, made ready for the next rover to be sent there by 2020. Maybe on next mission such a problem would be resolved instantly and with no two-week interruption of science while the engineers fix the problem.

A very big learning curve in store for inter-stellar missions. Maybe such missions could, as a last resort, always hold on to the capability to beam back home for any suggestions! But only as a last resort.

JoeP March 18, 2013 at 17:08

Ron has it right. This is more like adaptive circuits that have redundant components built-in, with programming to detect and reroute around damaged parts. You have paid a price however, in that you have less redundancy levels now, or an impact to performance or capability.

A step beyond the above is the ability of systems to make replacement parts as well. 3D printing and robotics will likely play a role, along with automated semiconductor manufacture. If components may be 100% recycled, then all you need (mostly) is sufficient power to keep the system working almost indefinitely, assuming nothing damages the system beyond recovery in a catastrophe.

RonSmith March 18, 2013 at 17:17

But wait, doesn’t this mean that when the Robot Revolt (TM) happens (aka Singularity V3) that the robots can not be stopped even with lasers? Perhaps this is not such a good idea after all. ;)

David A. Czuba March 18, 2013 at 18:22

Jesus’s maxim comes to mind (paraphrased): “Hardware, heal thyself!” Not only hardware, but software can be made self-healing as well. Ron has a good point, too, because compensating for loss is the constant companion of every journeying explorer, robot or human. Our desktop software might not be able to heal itself if several major .DLL or .CAB files go missing or damaged. If you’ve ever had a coworker move an important network folder inadvertently, you know what I mean. Humans shouldn’t have to run diagnostics like I did recently to defragment a virtual machine environment, even requiring me to pull out Microsoft’s Contig.exe application to defrag the pernicious Outlook.pst file. With solid state computer components that do conform to the Turing configuration, memory and instructions can be partitioned on any portion of the hardware. As for the non-Turing computing configurations, like those in use at Sandia Labs, systems that are resistant to intentional damage are also being developed. Again, if you’ve ever had a disgruntled employee nix not only data, but executables on the network, you know what I mean.

Eniac March 18, 2013 at 21:59

Because the maintenance of machinery, the replacement of parts, is inherently a mechanical process, a true self-healing machine will absolutely require mechanical manipulators. So, wardens are exactly the right idea. Self maintaining spaceships (or factories, or whatever) will always be crawling with robots agile enough to reach and replace every single part in the system. What is not needed is intelligence, artificial or otherwise. A sufficiently detailed set of instructions (aka genome) can operate such a system just fine. Living organisms do it, machines can, too.

Michael, March 19, 2013 at 2:14

I suppose it is more like re-routing than self-healing, self-healing in the general sense would be for the damaged silicon to regrow or reform, T1000 style -it does not look like much regrowth to me but maybe we are asking to much after it was hit by a gigga watt laser!

Alejandro A.Perez Rios March 19, 2013 at 2:41

The concept of self healing technology is a great initiative. By defining the parameters of the term self healing the technology can be applied to a miriad of practical applications. Artificial self healing does not have to be exactly nor similar to biological healing.

ljk March 19, 2013 at 10:25

Why not skip trying to reinvent the wheel with mechanical devices and go with a system that has been perfecting itself for over 4 billion years on this planet by making a bioship.

http://en.wikipedia.org/wiki/Bioship

Rob Henry March 19, 2013 at 16:52

Ron is right that “self healing” is an abysmal description of the hardware reroute used here. For that reason, I was wondering if this has a software only description, and thus if damaged hardware is red herring.

Would it still work if new components were ADDED to it in a slightly different area than lost components after multiple program reroutes had already been enacted. IF so, then this would be the real software component of a true self healing system.

GaryChurch March 19, 2013 at 18:50

“What is not needed is intelligence, artificial or otherwise.”

Having to troubleshoot a helicopter autopilot system would change your mind about that. Electro-mechanical hydraulic feedback systems defy any automated failure detection. I know.

There is no substitute for a technician; I have been hindered and seldom helped by many aircraft self-checking systems; none of them worked the way they were supposed to.

Comments on this entry are closed.

{ 1 trackback }