An even cooler project would be proving once and for all that a swarm of small, rugged drones can indeed cover a larger area in harsh conditions and are a much more sensible strategy in exploration than a large single drone.
Iff you have ten, twenty expendable machines then you don't need the same level of QA, thus opening up a lot of interesting possibilities. If we can get equivalent amounts of exploration done in just half (or maybe quarter) the cost then it opens up room for an arms race of sorts where each generation can be rapidly iterated upon and we can plan such missions within a one or two year timeframe instead of an eight year one.
Well, surely for a certain type of scientific mission (e.g. something like spectral analysis via ChemCam) you cannot just shrink down the components and instruments indefinitely. The size of Curiosity is directly related to the fact that it is bristling with scientific instruments like no other probe before it. Sure you could make a swarm of toaster sized drones that just go off and cover ground, but the purpose of this mission is not just to take some pictures but to do some very detailed scientific analysis.
What you could do is have dedicated drones which are essentially movable instruments, so you can have co-operation on tasks and say teams of 3 or 4 (one vaporises, the other has the spectrometer. One drills, the other has the reagents and so on)
If you enter this paradigm then you can eliminate costs such as that long arm which you need to position the instrument package. That arm in of itself is an engineering marvel and it requires a lot of careful design to make sure that it doesn't malfunction. (remember you have a not so light weight at the end, the torque due to that is huge and you have a complex assembly consisting linkages to transfer torque and so on...)
The idea of this is to see how simple and redundant you can make things. If for the cost of that arm we could have one small team of rovers wouldn't it be worth it? Wouldn't it jump start exploration?
What you might "gain" in reduced complexity by getting rid of the arm you loose massively in being able to power the whole system efficiently in one go. If you split the 10.6 pounds of plutonium across N smaller drones you would need to duplicate N Radioisotope Thermoelectric Generators (RTPs), power electronics, batteries, radios, cameras, computers and probably a half a dozen other things I haven't thought of yet. This probe is going to be going for at very least 10 years - probably quite a lot longer, all because of the design choices around how it's powered.
Again do we need all of those systems? Why can't you have two dedicated machines for earth communication and the rest communicate via a HF radio protocol? You don't need a RTG if it's say sojourner size and packs one payload on a standardised mobile base, solar panels will work just fine. Since you have redundancy built into the swarm you can skimp on a lot of things including computing power. Remember you can use group decision making for more complex "choices" and if you lose a few drones due to unforeseen obstacles, then objectively it shouldn't matter, so you don't need a lot of LIDAR and stuff as well. As far as heating and shielding goes then the smaller size actually works to your advantage.
I don't see any reason why someone shouldn't make this. Yes these probes will be disposable, but that's the entire point. They can be used and thrown away opening doors to risk taking that we haven't really seen before.
"I don't see any reason why someone shouldn't make this."
You are exactly the kind of person that should build your own lander if you don't see this. And please, stop asking all of these questions in bad faith. I applaud deeringc for engaging with your specific points, but I can't bring myself to do this, since it seems to me like you assume all aerospace engineers are uncreative drones that can't think outside the box and see the obvious solution you've reached from your armchair.
Have you ever built a multiply-redundant, space-worthy, swarm-based system before? Have you built something that's any one of those three? If not, I don't have a problem with you thinking about them, but I do have a problem with your attitude. Edit: to clarify, I mean the attitude that comes across from your writing style. I have no idea what your qualificaitons are, how much thought you've put into this, how receptive you are to the idea that you're wrong, whether or not you recognize that everyone has more "unknown unkowns" than anything else, or what your actual attitude is. All I have are the words you write here. And my natural response to snark is more snark.
And if you think you're the first person to think "wouldn't it be great if not every spacecraft didn't have to have to re-solve the problem of power, communications, and computation?" I first heard the idea proposed in a 2009 talk[1] by someone who has been an insider for over 20 years. Some gems I remember from that talk:
* Having an identical copy is not redundancy
* Complexity is inherently more succeptable to failures
* Cars have been widely successful because of gas stations and repair shops. Spacecraft have to drag around their own
* We could create a ring of satellites, each dedicated to providing comms to earth, wireless power, computing power, or whatever. Users of the system would just need to build a structure, interfacing components, and their instrument and launch it into a nearby orbit.
> You are exactly the kind of person that should build your own lander if you don't see this.
Once I have the financial resources, I actually plan to do so.
> Have you built something that's any one of those three?
To answer that yes I have, which is a completely orthogonal fact to my original comment because I am advocating reducing the qualifications of "space-worthiness" through the use of multiple copies. After reading your comment I decided that there had to be some respectable source who had advocated this earlier at some point and with some research I found this paper written by Rodney Brooks called Fast, Cheap and Out of Control : A robot invasion of the solar system outlining the same concept; http://people.csail.mit.edu/brooks/papers/fast-cheap.pdf
> I mean the attitude that comes across from your writing style.
Yes, I agree that my comments could look snarky under the weight of your assumptions, but I was doing my best to be genuine and engage in an honest discussion
> I have no idea what your qualificaitons are, how much thought you've put into this, how receptive you are to the idea that you're wrong
I'm actually quite certain that I'm wrong most of the times, but I don't know how I'm wrong and discussing, building things are the only ways to find out.
> whether or not you recognize that everyone has more "unknown unkowns" than anything else
Yes I do and it is a terrifying thought.
> or what your actual attitude is.
I'm doing my best to learn as much as I can and to never judge. (judgement takes up too many mental resources)
> And if you think you're the first person to think "wouldn't it be great if not every spacecraft didn't have to have to re-solve the problem of power, communications, and computation?"
I'm not and I would love to do more than just think and actually build things.
> * Having an identical copy is not redundancy
Can you please explain why? Is it because the failure points remain the same?
Hey, I'm glad you responded, and that I misinterpreted where you were coming from and your intentions, but I guess that happens when limited to text (not that a misunderstanding-free medium exists). But as you've probably gathered by now, most of my response came from reading "I don't see any reason why someone shouldn't make this. [and I doubt a good one exists]" instead of "I don't see any reason why someone shouldn't make this. [and I would like to hear some]"
> Fast, Cheap, and Out of Control
The first thing that popped into my head when I saw this title was NASA's "Faster, Better, Cheaper" initiative. I will admit that I do not know much about it, but what I do know is that there were some failures (as you would expect) and the public did not react well. The failures did not come from not having space-grade parts, IIRC, but for various other reasons. The most infamous was the Mars Climate Orbiter, known for failing due to an imperial-metric conversion error [1], that is still brought up in almost every space-exploration piece that gets sufficient attention. Anecdotally, the lessons learned by those I've talked to are (1) choose 2 and (2) the public will not tolerate failure on large NASA projects, even if those projects cost a fraction of what it takes to host an Olympics, or to buy Instagram, or fight a war for a day.[2] But I have just now hopefully started a discussion with my coworkers about this paper[3] on our message board.
So anyways, about the paper. It looks like your idea is only mentioned briefly in the one section, and not fleshed out. The idea was somewhat more fleshed out in the talk I went to. What I meant by my line of questioning was that I haven't seen an implementation of these concepts outside academia, even though the idea has been around for some time, and the industry isn't entirely driven by irrational beings, so there must be some technical reasons why the ideas haven't been fully adopted.
>> * Having an identical copy is not redundancy
> Is it because the failure points remain the same?
Essentially, yes. His argument that most failures were systematic, and not due to unequal 'wear' of various kinds. Software especially, since on space missions it is implemented to be as deterministic as possible. That means that even if the primary processor fails gracefully due to bad internal logic, and a hot backup immediately takes its place, the backup will behave in exactly the same way given those inputs, which are likely to stay the roughly the same across the switch. Another example is a mission where the high gain antenna succumbed to a systematic failure, and they completed the mission on its low gain antenna. If there had been 2 identical antennas, they would not have been able to.
Designing and testing components to last a long time under the conditions you expect and test for is relatively easy. It's when the designs and tests don't match reality is when the problems happen. If you write something and make a copy, it will retain all of the typose of the original, and it's the same with Code/CAD/etc (and you might be more likely to introduce bugs than fix them). Ground-based systems can get away with this because of how easy it is to replace/repair broken units in a redundant setup.
[1] Even though it was due to a much larger issue related to how the project was carried out, and just happened to manifest in that error. It could have easily been a kilometer/meter mixup instead. And having a backup string of landing hardware/software wouldn't have helped in this case, unless the backup units had a different design/implementation.
[2] And yet, JPL still chose to use the skycrane. Very couragous.
>>> most of my response came from reading "I don't see any reason why someone shouldn't make this. [and I doubt a good one exists]" instead of "I don't see any reason why someone shouldn't make this. [and I would like to hear some]"<<<
Ah I see. I'll start qualifying my I-don't-see-any-reason-whys from now on with the latter statement. :)
>>> I will admit that I do not know much about it, but what I do know is that there were some failures (as you would expect) and the public did not react well.<<<
I guess that's the real reason why this won't be implemented by government agencies. Space missions are a matter of national pride and no one wants a "designed to fail/waste of money" accusation on their hands. However, empirically speaking it seems to be the best way to do things as we can increase the amount of explored area quite rapidly and respond to changes much more quickly. I think that when private companies such as Planetary Resources start doing exploration they will be forced to adopt this model because of their constraints and a lot of amazing solutions will come out of it. If that happens then it might open up a pandora's box and advances will happen much more quickly. Some people will see it as a bad thing, but as long as the systems are autonomous it should lead to a lot of good things.
>>> 2) the public will not tolerate failure on large NASA projects, even if those projects cost a fraction of what it takes to host an Olympics, or to buy Instagram, or fight a war for a day.<<<
More importantly the politician who approves it probably won't get re-elected.
>>> [2] And yet, JPL still chose to use the skycrane. Very courageous.<<<
I was quite shocked when I heard it worked. I was willing to bet on the side of failure because of the sheer complexity involved. A small timing error, sensor glitch or the other million things that could have gone wrong would have led to failure. It's quite impressive that they managed to do such a high stakes real-time task more or less autonomously. It really was quite a daring thing.
>>>your idea<<<
It's definitely not mine. I bet people were talking about it when I was in diapers.
>>> What I meant by my line of questioning was that I haven't seen an implementation of these concepts outside academia, even though the idea has been around for some time, and the industry isn't entirely driven by irrational beings, so there must be some technical reasons why the ideas haven't been fully adopted.<<<
Yes, I'm willing to bet that this has a lethal flaw which quickly led to such implementations to be rejected, but the question is can this be hacked, for the lack of a better word? Remember most organisations where rovers are designed are meant to be risk averse and they, aside from NASA, have an endless pool of resources to draw from. (I'm talking about the military) The sociological and resource incentives in place simply work against any such proposal independent of engineering viability.
>>> His argument that most failures were systematic, and not due to unequal 'wear' of various kinds. <<<
There's an interesting case to be made over here, one is redundancy at the unit level and another is redundancy at the systemic level. Let's take the swarm as a system, in that case if you have multiple backup copies of machines adept at performing particular tasks then you essentially have redundancy provided that the same method is not followed at the unit level. In that case if one failed for a particular set of inputs (hardware or software) then you can "patch" the rest by either avoiding the physical situation or changing the software. You can repeat that with individual units and let them have free reign. If one of them gets destroyed then the other ought to be able to be modified in time. Although it won't guard against stupidity at the unit level, this should to be a much more redundant system than anything we can create within a single one-use device.
At the level of the single unit, systematic failures come into play and more copies is actually less over there. Things must be asymmetrically designed to overcome edge cases and systematic failure modes and robust engineering makes sense over there. I think that one of the best ways to implement such a system would be to spend most of the resources in creating and testing a mobile base which is independent of its payload. (attachable modules which perform specific tasks) You should then be able to deploy this machine across missions and learn from all of the real world testing in each mission to create something truly robust and reliable. Once you achieve that you can start offering redundancy at the swarm level through the payloads. For example in that high gain, low gain antenna scenario, wouldn't it be better to have a dozen robots equipped with a variety of antennas dedicated to communication?
>>> It's when the designs and tests don't match reality is when the problems happen. <<<
Yes, but isn't the entire point of the exercise to fail early and fail often so that you can succeed? If your system is disposable then any systemic failure becomes yet another data point to engineer against and all future systems are better because of it. There is no better laboratory than nature and surely this is a point in favour of it? (unless I'm missing something)
Would you like to carry out this convo via email? If so then please feel free to drop me a line at searchingforabsolution [at] hush.com
Yes. Like I've said above the explicit purpose of this mission is to gather advanced scientific data which requires comparatively bulky instruments and high power availability. Because it is so ridiculously expensive to get anything to Mars you want it to last as long as possible. "Disposable drones" make absolutely no sense when it costs so much to launch them into orbit, transport them 60 million km across the solar system, and then enter the Martian atmosphere and land in a coordinated place on the surface. Each gram you get to that point costs many thousand dollars. You don't just plan your strategy around losing a bunch of them - that would be many hundreds of millions of dollars down the drain.
Using a decaying nuclear isotope to power your probe you will get many multiples more bang for buck compared to solar powered probes which so far have maxed out at about 5 years. The solar concentration of a Martian winter is extremely low, and this problem is exasperated the smaller the probe and the resulting battery that it can carry.
Just like in computing and electronics, physical and mechanical distributed systems are inherently complex - more so than monolithic systems.
>>> Yes. Like I've said above the explicit purpose of this mission is to gather advanced scientific data which requires comparatively bulky instruments and high power availability. Because it is so ridiculously expensive to get anything to Mars you want it to last as long as possible. "Disposable drones" make absolutely no sense when it costs so much to launch them into orbit, transport them 60 million km across the solar system, and then enter the Martian atmosphere and land in a coordinated place on the surface. Each gram you get to that point costs many thousand dollars. You don't just plan your strategy around losing a bunch of them - that would be many hundreds of millions of dollars down the drain. <<<
I've thought about what you have said and I think that we are measuring the likelihood of success in different ways. I'm measuring it in terms of the likelihood one pair of devices will complete the outcome at the cost of all the others and, if I am correct, at some level you are measuring it in terms of reducing the possibility of a loss while achieving the mission objectives.
I think that we can afford to build disposable machines because if they are tiny and can fit within say a 50 cm cube (which is the diameter of curiosity's wheel) the mass of each machine will also be radically less. Curiosity weighs 899 KG, a well designed vehicle base could weigh as less as 1 KG with instrumentation we could work with the assumption of 2 KG. That is around 450 rovers! If they are divided into teams of say 6 and are dropped off using some method at discrete intervals then you have 75 teams exploring the martian surface. If each team explores during just the warm martian months (I'm working with assumption of 400 sols) with a rate of a very conservative .5 m^2 explored in a day then all of the teams combined will explore 15000 m^2 in the course of a single mission. That's huge. Further, in this scenario, if individual units fail at some point then the entire mission won't be jeopardised and that number will roughly stay the same. I think that if the units are allowed to be autonomous (again because they are disposable) then you could rapidly increase the area explored and get more out of a single mission.
In this scenario the success of the mission has now bifurcated from the functioning of a single device and because of that you are free to pursue several orthogonal benefits such as these which ultimately reduce costs. I think that if you factor in a decrease of launch costs due to company's like SpaceX, then this ought to become even more attractive.
>>> The solar concentration of a Martian winter is extremely low, and this problem is exasperated the smaller the probe and the resulting battery that it can carry. <<<
One of the main uses of the battery during winter months is to keep the processor warm. If the assembly is small enough then you should be able to use just insulation, a very small heater and perhaps a long lasting exothermic reaction which proceeds slowly over time. The amount of heat generated by such a reaction would be too small for something like curiosity, but perhaps it might work for a very small machine? Again since tolerances are low you shouldn't you be able to use a wider variety of batteries which store more per unit volume? I might be wrong on all counts, but a smaller design and lowered tolerances might actually work to our advantage.
>>> Just like in computing and electronics, physical and mechanical distributed systems are inherently complex - more so than monolithic systems. <<<
I'm actually not that into computing and electronics, I used to build physical systems and how they fail fascinates me. My designs failed so often upon meeting the real world that I realised the only way to know if something would ever work was to actually implement it IRL. If you can carry out a mission at one-tenth of the cost then you can do that much more willingly and learn from unforeseeable failure modes much more quickly. It should be an answer to this problem than the other way around.
There is at least MetNet (http://en.wikipedia.org/wiki/MetNet) which aims to swarm of measurement devices on Mars surface. They are going to use immobile measurement platforms.
Iff you have ten, twenty expendable machines then you don't need the same level of QA, thus opening up a lot of interesting possibilities. If we can get equivalent amounts of exploration done in just half (or maybe quarter) the cost then it opens up room for an arms race of sorts where each generation can be rapidly iterated upon and we can plan such missions within a one or two year timeframe instead of an eight year one.
Of course, this is easier said than done... ;)