Electronics of the shelve get very unreliable in high dosage environments.
The radiation rewrites your loaded programs, by flipping bits and causing rising edges where there shouldn't be some.
So the program running on your camera will dissolve above some radiation levels, locking it in a watchdog detects and reboot cycle. There are approaches to circumvent that in low to medium environment- one is to have redundant systems and select the "majority" vote on computation results.
But at this level, the only solution is to not have computation near the source and put up with long cable/fibre cable delay- which is tough with complex sensors (aka computers) like cameras.
It is probably some snail-eye setup- the camera a remote mini-drone on cable, retracted on black out.
In my day-job, I (help) design electronics that is subjected to (much less) radiation, and we also irradiate our devices to check for long-term effects (and this is done with a comparable doserate in a facility for this purpose).
Damage to your electronics in practice is almost exclusively caused by Photons (convention is to call this γ-Radiation when caused by radioactive decay in the nucleus, X-Rays when created in an accelerator). β-radiation (fast electrons) is easily shielded by thin layers of metal, and α (He nuclei) can't penetrate sheets of paper.
Individual photons can't really deposit much energy at a single location in your semiconductor, so they aren't able to generate enough charge to "flip bits" instantly. Flash still uses a lot of charge/bit, so it's relatively stable, DRAM is constantly refreshed and SRAM would need a jolt of high current to flip, so that does not really happen, either.
What radiation does, however, is to slowly damage the Silicon and change its crystal structure (introducing defects) which increases leakage and moves the analog threshold voltages of circuits around in a funny way. So, what we often see is flash becoming un-programmable on a more global scale (rather than individual stuck bits), and most importantly the analog aspects (voltage references, brownout-protection circuits, voltage regulators) cease to function.
This is all very variable, but we normally observe effects starting at "a few 100 Gy" when testing more complicated modules. We don't research individual components, though.
One especially nasty type or radiation are neutrons, though. When being "moderated" (slowly decelerated by successive interactions with material) they tend to have a high likelihood of merging with some other nucleus (being captured), and the energy resulting from this capture effect can be huge and concentrated on a single spot. This can be enough to flip bits, and this indeed may be an issue near the funny isotope mixture present in Fokushima at various places. Having a strong neutron emitter is very uncommon, though.
Unfortunately these damages are not specific to digital electronics, and especially optical components (parent refers to long fibre cables) tend to be rather sensitive due to their large structures, so "keeping the computers out" may not help much overall.
I dimly remember reading about a chemical treatment to the material, that would react with the ionized silicon until the reaction was depleted.
I do not remember the substance- and it was ridiculous expensive (in addition to the ridiculous expenses of custom made chips)
So, simple question: why isn't everything electronic shielded inside big lead blocks?
I get why this isn't the prevailing approach in aerospace applications: weight. But for a ground based robot with an external power source, why not heavy shielding cubes with minimal connections to the necessary exposed bits?
And why not just load it down with 5x CotS sacrificial cameras, then expose them as needed when the previous one dies?
For typical Gamma radiation of about 1MeV photon energy or higher, you need about 1cm thickness of lead to reduce the radiation to about a third (by a factor of 1/ℯ). One order of magnitude of radiation hardness (reduce radiation to 1/10th) needs two centimeters of lead. That's getting heavy pretty fast.
Radiation intensity inside your box depends on the thickness t: I(t) = I₀·exp(-tρμ)
t: thickness, say 1 cm
ρ: density of the material, for Lead 11 g/cm³
μ: absorption coeffcient 0.1 cm²/g [see ref 1]
I₀: intensity outside of the box
...import numpy as np...
In [5]: np.exp(-0.1 * 11.34)
Out[5]: 0.32174370422037013
In this case there is zero need for electronics on the robot. You can have a pure mechanical system with a fiber optic camera etc like a endoscope. Power as torc through a plumbers snake. It can then either carry a box with some sensor package, or take remote samples. For extreme radiation, you can put a Film badge dosimeter in a water tub to work out the radiation levels indirectly.
It means, that the robots have to be very "custom" build ones. With the intelligent parts beeing deposited outside of the dangerous area, communicating with the muscle via e.g. fibreoptics.
The robot consists ideally of some die-hard sensors positioned as far away as possible and as close as necessary - which are retractable mechanically in case of software failure by left behind elements. Then there is the actual tooling, which is basically just a remote controlled shell. A complicated setup, to say the least.
Imagine a Anemone with some little remote camera fishes on a tentacle, controlling a robotic crab on the longest tentacle of them all.
Ideally you'd want it to be gamma-transparent as well; otherwise, those gamma photons which hit it are going to deposit their energy into it. This is how glass viewing windows in nuclear materials processing facilities, for example, gradually cloud and embrittle over time, and have to be replaced. I'm not sure how well a relatively thin optical fiber would stand up, especially under the kind of bombardment we're talking about here.
You can stop them by slowing them down. Usually you'd use ellastic collisions with light atoms like hydrogen. Water works which is what reactors use. For shipping polyethylene is often used. Then some materials simply have a higher cross section for absorbing the slow ones. Had to look this up I saw boron and cadmium.
(Married to person who used to ship Californium-252)
You can use humans to clean the mess, whose massively redundant computational model is much more robust to radiation damage. They will however suffer or die from radiation exposure long term.
This is what was done at Chernobyl in admittedly much more dire situation. A quantity of Japanese remote-controlled robots were urgently imported back then as well, and failed in the same manner.
Not this mess, or even that one, really. The "bio-robots", so called by their masters, were mainly collecting debris from the reactor explosion for containment; they were not trying to deal with the core slag sunk into the basement of the building, which at that time was still thermally very hot as well as being ferociously active. Even to approach that would have been immediately lethal.
Given the cited dose rate, the same is true at Fukushima - an acute dose of 10Sv is very likely lethal on its own, and at 650Sv/h you get more than that every minute. Anyone you send into that is going to be unable to work within seconds, and dead inside a couple of minutes tops.
That's good to know, thanks. On the other hand, I find sources calling that dose rate lethal within minutes, and at least some of the Fukushima corium is roughly seven times as active. Even if you're willing to write off as many lives as it takes to reduce what is essentially solidified lava by means of hand tools into a containable form, I doubt it's likely to succeed simply because it'll kill off your workers so fast that, after a little while, no further progress is possible because of all the corpses blocking the way.
Well I mentioned it just as an example of humans being better at it than robots. I don't really imply that sending people to death is a good idea.
We should also distinguish between site cleanup (which is certainly impossible) and secure containment (which is possible and was done before). The idea is not to chip away bits of highly radioactive substance, but find a safe way around it to isolate the world from further contamination. Apparently, not everyone is convinced Japan has a plan there.
Realize that the ostensible safety of nuclear energy is based on believing that black swan environmental and political events are impossible. Walk away from it and move heavily/quickly toward renewables.
So the program running on your camera will dissolve above some radiation levels, locking it in a watchdog detects and reboot cycle. There are approaches to circumvent that in low to medium environment- one is to have redundant systems and select the "majority" vote on computation results.
But at this level, the only solution is to not have computation near the source and put up with long cable/fibre cable delay- which is tough with complex sensors (aka computers) like cameras.
It is probably some snail-eye setup- the camera a remote mini-drone on cable, retracted on black out.