Marine Corps Shelves Futuristic Robo-Mule Due to Noise Concerns
military.comFor those who don't understand the U.S. military funding process: This is being shelved because there isn't money to continue work, not because of a fundamental problem with the technology. [1] It's more of a "We'd LOVE to give you this cool thing... Too bad we don't have any money to work on it. Oh well!"
[1] The article mentions that they built a quieter version of the robot that carried less. More interesting were problems about how to maintain the things (from training mechanics to supplying replacement parts) and how to integrate them into current training. (How/when to use? Advantages? Disadvantages?)
[2] For those of you interested in the U.S. Military's use of Mules in more recent times: http://www.newyorker.com/magazine/2010/02/15/riding-high
BigDog's power pack was originally a constant speed gasoline engine driving a hydraulic pump and an electrical generator. That's the noisy engine in the earlier videos. The LS3 was supposed to have a new, quieter power pack developed with a small variable-speed Diesel engine. That was subcontracted out by Boston Dynamics. This is as good as it ever got: [1] It still sounds like a dirt bike.
The hydraulic system used is very controllable but not particularly efficient. There's no energy recovery and no springyness; it's brute-force hydraulics. That was reasonable for an experimental machine, but not acceptable in the production product. Atlas, the BD humanoid, has the same problem. It's too similar to BigDog, and weighs about 330 pounds. Schaft, Google's other humanoid robotics company, uses water-cooled electric motors, like Tesla. You can enormously overload electric motors for a few seconds without hurting them, and if you have cooling and temperature monitoring, that works fine. This is probably the way forward for anything smaller than a pony.
One big drawback not called out to electric is you have to power it to have it hold still. Hydraulic can just leave a valve closed which may or may not require power.
Then again just have it "sit". This may not be the best tactically though in small group engagements as "sit" is movement and freeze is dead stop.
I'll let people who know more about that debate it though.
> you have to power it to have it hold still
Is it common to work around this by supplementing the electric motor with a disc brake?
Depending where you want to be on the tradeoff between complexity vs energy-efficiency, you could use a hydraulically-powered disc brake in order for the brake to not need power except during state transitions.
Thought about this for a bit. I'm betting it's used. The deal with brakes is the ratio of the friction applicable by the brake and the lever distance of the compared to the lever arm it's stopping. I'm watching a movie so I'm not going to do the math right now but I think the size of disc rotor you could get on say a mule's knee joint versus the lever arm length of say a lower leg might cause problems either with the size of the rotor.
My thinking is mountain bike discs are 120mm/4.8"-203mm/8" diameter compared to a 29" wheel.
These are meant to stop a 160lb person who is going down a hill. Well Slow to a stop not lock dead in place as the wheel would break free so it's not the full stopping power of a real.
Anyway my gut reaction is it might be difficult in this use case. I think a 8" rotor on a similar lever arm might have trouble fully locking down that joint and would be very large and flimsy (8" thin sheet of metal) for rugged use. Plausible though.
But hey I failed out of mechanical engineering into computer science for a reason so maybe I shouldn't be trusted :)
Many robotics applications use ballscrew linear actuators. These contain a cage of ball bearings in a spiral around a ball screw, and they move a given distance for each revolution. You can apply a braking force to the ball screw, which will have a large mechanical advantage over trying to brake against the joint itself.
Very true. My cnc uses these. However I couldn't see getting it to walk or hop like these mules. You don't really get impulse with a ballscrew.
Think smaller. Most of the mechanism has to fit inside the dimensions of an arm or elbow.
I think you're missing that this is EXACTLY my point. I'm really doubting a disc that could fit inside an elbow would be able to stop the torque of a few hundred (900/4) lbs on it. Actually one inside the elbow would have a very minimal surface area on the disc in the first place.
To complement other answers, it seems a 3-phase AC induction motor has good torque at w=0. Also, in theory it's power consumption should be 0, but of course there's the non-negligible resistance of the winding and other losses, but the EMF will still be small, so P=VI is small.
> One big drawback not called out to electric is you have to power it to have it hold still.
How so? There are plenty of ways to mechanically lock a joint in place.
Sure there are. They add complexity and weight and failure conditions. There's also a difference between lock and place and lock at an exact position.
That may be, but the application never really seemed reasonable. I say this as someone who loves almost anything robotic.
The technology is amazing and will no doubt find use elsewhere.
These are the "Big Dog" mules being built by Boston Dynamics, who was recently acquired by Google.
Google has been against militarizing the robots since acquisition, so they're no doubt happy to have this contract dropped.
To be clear, this doesn't mean the project is shelved, it means the contract with the Marine Corps is shelved. Google will be able to focus more directly on civilian uses for the mules from here on out.
Well this technology certainly has far more uses than the military, specifically search and rescue. Send these puppies in to hard hit areas, isolated spots, and the like, with supplies and such. The military will get their robots so we have to accept that but we can use all that research into making them do useful and humanitarian missions. No one will care if that rescue robot is noisy.
In general the amount of U.S. government spending for "search and rescue" is dwarfed by spending on the military. In that context having a military application (or the pretense of one) for this tech gives it access to the money necessary to continue research.
Why are they against militarizing robots? It pays the bills. They could start with consumers or business but it's harder to build a market. This makes it harder to fund further development.
Once the technology is developed by anyone for non-military use, transitioning to the military is easy. There's not some magic that prevents consumer developed technology from reaching the battlefield.
I think it's just a principle thing. The founders built Google on a policy of "Don't be evil", and war-robots may be considered evil depending on your own perspective.
My point is that you are still building war-robots, you just aren't getting paid for it. The military will just pay someone else to build the military version.
In 2015, we made great progress in identifying a person with a gun:
http://www.bidnessetc.com/59108-microsoft-corporation-beats-...
Bet this technology soon makes it into drones.
> Why are they against militarizing robots? It pays the bills.
Because money isn't everything?
> There's not some magic that prevents consumer developed technology from reaching the battlefield.
That doesn't mean you're required to actively help it along.
You are actively helping it along, you simply lack the ability to connect the dots.
Over the next 10 years, the military should fund building World Cup robotic soccer teams so people such as yourself can feel comfortable with the non-military technology. In year 11, they can add the weapon.
Google is based on the internet which came out of DOD research
And the nature of DoD research back then was more generally applicable to things besides killing. The Mansfield Amendment is what turned ARPA into DARPA and pushed to curtail pure research. If you couldn't kill someone with it, it wouldn't get funded.
The Mansfield Amendment of 1973 expressly limited appropriations for defense research (through ARPA/DARPA) only to projects with direct military application. Some[who?] contend that the amendment devastated American science, since ARPA/DARPA was a major funding source for basic science projects of the time; the National Science Foundation never made up the difference as expected. https://en.wikipedia.org/wiki/DARPA#Later_historyThe Internet didn't only come out of military research and spending. A lot of the technology, thinking, and origination came out of the private sector and universities as well. Not to mention the actual early days of the Internet were overwhelmingly private + university based, and not military. Nearly all of the advances that the Web was actually built upon, were not military. The deployment of Google occurred on top of an almost entirely private infrastructure circa the late 1990s. The specific technology stack that made Google possible at the time was almost entirely not military.
There are these things called "mules." They're quiet, can carry about 200lbs., and can refuel themselves as they go by eating grass and drinking water...
The logistics of keeping them fed and watered, having trained handlers, and treating them for injuries are nontrivial. And they're prone to unhelpful reactions in a firefight. There are good reasons they've largely been displaced by motor vehicles. Combining the go-anywhere ability of legged propulsion with the advantages of motor transport seems like a good enough idea to merit study.
Much easier than the logistics of keeping a robot operational in the field. You don't care whether the Taliban captures your flesh and blood mule, but you really don't want them capturing your world's-most-advanced robotic mule.
The Taliban are not going to have the capability to reverse engineer and produce their own copy, let along having the logistic capability of keeping it in the field.
But the parties they sell it to, for substantial combat-enabling cash, will.
I'm sure they could find a buyer for it that can.
Is it common for military electronics to have a self-destruct feature to protect technology secrets?
E.g., when damaged beyond repair or abandoned in a hasty retreat, the hardware could ignite a little thermite to incinerate the electronics boards.
One of the stealth helicopters used in the assassination of bin Laden was so destroyed, but a part of the tail was left behind.
https://en.wikipedia.org/wiki/Death_of_Osama_bin_Laden#Helic...
I used to have some old (tube-based) test equipment that included instructions on exploding it with a grenade in the event that it had to be left behind. The grenade was not built in, however.
Why? What are they going to do with it?
Sell it to a country that has the technology level needed to copy it. Russia and China likely would be interested; they might not want to do business with the Taliban directly, but might be willing to buy them through a middle man, say Pakistan.
It is fairly common for countries to let other powers take a look at technology that lands in their hands (sometimes before returning it to them). See for example https://en.m.wikipedia.org/wiki/Viktor_Belenko#Aftermath
A lot of the ability of the Boston Robotics robots is in their software thought which is much easier to lockdown via encryption. The rest is the hydraulic mechanism which is not easy but is the best known part of what the BR robots do.
Concealing algorithms and executable code is actually very difficult, consider that to run, a program must first be decrypted. Also consider the decades long war between crackers and proprietary software creators. Obfuscating algorithms has been the subject of quite a lot of academic research.
Store the decryption key in a removable module or the wireless controller? Then anyone who captures one is stuck with a ball of encrypted code that can't execute. Dealing with securing software and keys against capture is not new for the military.
Also destroying the hdd/ssd is easier than destroying the whole machine. You could even have a secure wipe option if the machine is still able to run it's computer but is disabled mechanically.
Yup, I'm sure the countermeasures you can think of in the time required to write an HN post will be secure against technically-advanced nations' militaries.
I was thinking more along the lines of thermite
Sure and it's a problem the military has been dealing with for decades. I'm going to assume they understand the calculus and some effective countermeasures.
Perhaps reverse engineer it, find someone that can reverse engineer it, exploit it or others like it, expose every detail about it all over the internet, a number of potentially harmful things?
And training a muleteer takes a long time also non mechanised units require a lot more logistic support.
The USMC small wars manual has a lot on how to organise an non mechanized unit that was mostly horse /mule and ox drawn.
Yeah, but they get all nervous around gunfire and explosions. They don't handle helicopters and the airdrop well either. Packing several dozen of them to fit into a C130 for the 12 hour overseas trip is also a bit difficult.
> and can refuel themselves as they go by eating grass and drinking water...
Try finding 30 litres of water and 8 kg of grass ( plus a portion of salt ) per mule per day, minimum, in the mountains of Afghanistan.
To move 90 kg of payload? There are many more efficient methods.
Actually they work great.
We (Air Force Special Operation) used them and horses to great effect in 2001 and the ANA have been using them before and ever since.
http://www.defensemedianetwork.com/stories/operation-endurin...
https://www.washingtonpost.com/world/asia_pacific/for-afghan...
>Try finding 30 litres of water and 8 kg of grass ( plus a portion of salt ) per mule per day, minimum, in the mountains of Afghanistan.
Is it any easier than finding diesel fuel?
Though sometimes the tables are turned: http://imgfave-herokuapp-com.global.ssl.fastly.net/image_cac...
And they don't cost $40+mm
And you can eat them in times of scarcity. The robotic ones aren't nearly as tasty.
Speak for yourself. Diets motivated by iron deficiency are a real thing.
Although in modern times it's almost ridiculously hard to avoid added iron in food these days due to enforced fortification based on outdated ideas from the 1940's.
per http://well.blogs.nytimes.com/2012/08/13/a-host-of-ills-when... , iron deficiency prevalence is 3% and iron excess intake prevalence is 12%.
> $40+mm
40 million million dollars?
I think most other people use a single M to mean million, and k for thousand.
Can't we just use SI prefixes or spell the numbers out instead of using some random alternate convention?
This is not a "random alternate convention", it's a highly used convention that you aren't familiar with. I wasn't familiar with it, either, until I moved to the Silicon Valley. Now I see it all the time.
It's a ridiculous "convention" and it's dubious what you mean by "highly used."
Using Roman numerals for anything is a waste of everyone's time.
There exists a world of finance outside of Silicon Valley. M and MM to indicate thousands and millions are a finance convention, not an SV one, people in SV just happen to talk about finance a lot. Anyone in finance would be familiar with it, and given that the amount here is a dollar amount, the use seems perfectly fair.
I strongly suspect that the page's etymology is wrong -- or more precisely -- not entirely correct. The Latin word for "thousand" is mille, which why the numeral for a thousand is M.
No one would be so asinine to mix numeral systems. Same thing with percent, which originally was rendered as "per-cent", as in "per hundred".
I don't think it's necessarily actually Roman numerals, but perhaps "mille", which means thousand. So thousand thousand means million.
It's "highly used" in a specific field and nowhere else. Outside of that field, you should use broadly-accepted conventions.
There is no Roman numeral m only M, so if they are sticking with convention shouldn't it be capitalised?
Probably because it's not a numeral, but rather an abbreviation for mille.
I completely agree with you, while I know $40+mm means $40M it is only because of context, not because I actually parse it as "thousand thousand". In my mind mm is millimeter, even as an American.
HN is the only place I see this notation used.
The great thing about standards is there are so many to choose from.
Yaks and camels too, depending on the environment.
And if you ever run out of things to do, you can shave your yak.
Can anyone in robotics weigh in on whether new approaches like pneubotics / otherlab [0, 1] make Boston Dynamics' super heavy/noisy/expensive robots irrelevant or not? That's what Saul Griffith has been saying at least [2], I'd like to hear an impartial take on it
[0] http://www.pneubotics.com/
[1] https://otherlab.com/projects
[2] "[...] we're the only game in town." 21:00 onwards https://youtu.be/gyMowPAJwqo?t=1263
Boston Dynamics' primary innovation is in the software/control design. The reason their robots are noisy is because they really haven't innovated a whole lot in the basic actuator design. That's not to say their mechanical design was easy, not by a long shot. It's super impressive stuff!
Any new inventions that would make their mechanical design steer away from loud hydraulics would no doubt be amazing. However, without Boston dynamics' delightfully ground-breaking controls software design, it'll be dead in the water!
For a little bit of proof, see who they are hiring! http://www.bostondynamics.com/bd_jobs.html
A lot (most?) of the noise comes from the engine, not the hydraulics, and the engine can't be replaced because sufficiently high-capacity batteries do not exist yet. That's not a problem that can be solved by innovative actuators.
That too. Also engine design was contracted out.
Thanks for your reply.
Is there much publicly available information about their software/control design innovations?
Hmm, maybe one might expect they would be hiring more mechatronics engineers to do just that? Or maybe there's still not much point for them because as you mentioned their focus is slightly different currently.
A person with a core Mechatronics focus will probably be a great systems integrator. BD's core workforce are these PhD's in controls (usually from ECE departments). Source: Am a Mechatronics Engineer with a Controls focus.
check out some of the work of BD's CEO (Marc Reibert). This (http://web.unair.ac.id/admin/file/f_7773_bigdog.pdf) is the summary paper on Big Dog, but they cite a bunch of the older Reibert papers on Controls designs for legs.
In my opinion, I don't think inflatable robots are very practical.(except for niche applications)
One of the most important things in robotics is reliability, the more work your robot does before breaking down or needing maintenance the more money it makes you. In industrial robot arms the gold standard for this is a Mean Time Between Failures of more than 100,000 hours. That's more than 10 years of continuous operation!
I am skeptical that inflatable robots will be able to last this long. The fabric/elastomer combo is certainly not going to last through 10 years of continuous operation. Sure this part may be cheap, but you still need expensive valving and pneumatics to control said robot. Not to mention that maintenance costs money too. The payback period for robots is also shortening,
The other problem that pneumatic robots suffer from is that air is compressible. This means that moving it around to drive pneumatics is gonna be inefficient and that pneumatic structures aren't that rigid. Higher rigidity means higher resonant frequency which means your robot can operate faster without wobbling around. The inefficiency might be compensated for because the robot is so light, but I have yet to see any hard numbers on this.
This lack of rigidity is touted as a feature by the people who make inflatable robots. Because they are so light and aren't rigid they aren't going to hurt people if they fuck up. There are other ways to solve this problem that are currently used in the robotics industry. One of them is to put a spring on every link in the robot, which is what the Baxter robot does. Another is to make the robot as light as possible and limit speed which is what one of Kuka's human safe robots does[0]. Better control also fixes this problem, if you don't hit the human then you don't have any problem.
But there might be niche applications. Maybe they will find use in the medical field where having anything rigid touch a human is unacceptable or where you need a weird shape to grab a human on a bed. Entertainment might be another, a while back a japanese company made giant inflatable robots for parades.
(http://www.phriends.eu/URAI_08.pdf)
That's a very good paper. There are many advantages to using two opposed springs driven by actuators to simulate muscles. You get muscle-like properties. You get energy storage and recovery. (Humans recover about 70% of energy from muscle springiness when running. Cheetahs, 90%. BigDog, 0%.)
As that paper points out, there are several ways to do this. The cleanest is a double-ended pneumatic cylinder with proportional spool valves at each end able to connect to pressure or exhaust. That was tried on a legged robot at CWRU some years ago. There are schemes with linear springs, string, two motors, and linkages, which tend to be bulky and complex.[1][2] Those work, but are more of a research design than a production mechanism. Somebody will do a better design, probably with rotational springs and no strings.
I once considered a design with two motors, rotational springs, and a differential. One motor controls impedance, the other controls position. If you don't need to change impedance rapidly, which you usually don't, the impedance motor can be much smaller and geared down.
[1] http://mech.vub.ac.be/multibody/topics/maccepa.htm [2] http://www.inacomm2013.ammindia.org/Papers/106-inacomm2013_s...
Hey thanks for responding!
If the parts are cheap enough, might it not matter that they don't last the regular MTTF?
One of the arguments I've seen them using is that their robots can actually move faster than regular ones due to their weight advantage - currently I would guess accuracy is not as good though.
Another point I've been thinking about is that biological systems tend to combine the rigid and non-rigid per the needs of the organism, right? It seems that combining these techniques could also be interesting.
why do they need this. What are they carrying? Are marines supposed to go without resupplies for weeks/months where they would need an additional robot mule to carry everything?
Combat loads have been increasing for decades as we load our troops with new tech and body armor. A US Army report indicates an average of 87-127 lbs (40-58 kg) carry load [0]. Another here [1] indicates 90-110 lbs. US Army doctrine prescribes a max load of 60lb, but troops regularly are exceeding the guidance due to operational necessity. It's incredibly stressful on the body and is a major indicator in early joint and soft tissue degeneration in our soldiers and Marines.
The branches are all looking at ways to reduce the load. Pack "animals" are an excellent option to offload heavy equipment and improve the health and long-term mobility of our combat forces.
0: http://archive.armytimes.com/article/20110214/NEWS/102140308... 1: http://www.globalsecurity.org/military/library/report/call/c...
An old coworker left the Army due to back issues from carrying gear through Afghanistan. Unfortunately their gear is often much heavier and lower quality than what civilians use for backpacking/camping.
Yes, actually this was a frequent occurrence in Afghanistan. Soldiers or Marines would be posted in small numbers in a remote COP in the mountains. Resupplying these troops would require more soldiers then were actually in the COP, and the resupply missions were often more dangerous than the mission of the COP itself (because it is easier to ambush a convoy of trucks than attacked an outpost).
Yes. In general the ratio is, I think, 10 support soldiers to one combat soldier. The logistics required to keep an army in the field half way around the world is enormous.> Resupplying these troops would require more soldiers > then were actually in the COP...
I found this interesting article:
http://www.nationaldefensemagazine.org/archive/2013/march/pa...
Which includes this quote:
"“What we’ve seen is that when Marines come ashore, they’re carrying 130 pounds of food, water, batteries, ammo, you name it, on their backs because fundamentally, they don’t trust sea-based logistics to keep them supplied,”
Which is ironic, because since World War 2 the Marines basically function as a de facto Army, and enjoy all the benefits of American global air and sea superiority.
Really, as a former soldier myself, the ideal solution is to stop having soldiers and Marines carrying so much damn gear. Much, much easier said than done.
*By ironic I mean "silly ass Marines"
> resupplies
Telling your robot to carry this shit over there is great way to resupply. Compared to having soldiers (or contractors) get blown up driving trucks around.
I wonder at what price point the civilian market could support these? There must be a fair number of applications for resupply where humans (sherpas), actual animal mules or helicopters are required, ie, forestry, National Park rangers, science, recreation, etc.
> forestry, National Park rangers
In most areas where these guys operate, actual mules are probably cheaper, and can be refueled with locally available forage. Not to mention they're more suitable for that kind of environment.
Perhaps more 'extreme' environments where that forage is not available would be more suitable?
Not arguing but I'm curious what the total cost of ownership of a mule is. Handlers, stabling, feed, medical care, waste, etc. And all that is on a year round basis, you can't just stick it in the garage in the off season. Not to mention a human required to accompany it to the destination.
Walking the wonderland trail a few years ago I was remarking to my friend that in some of the most remote meadows at the top (7-8k feet) (https://www.google.com/maps/@46.9181125,-121.7649511,731m/da...) of the trail there were long stretches of very nice chat trails. He was saying what a pain it would be to bring the chat up and was trying to decide if it was ATVs or mules or whatever. I guessed helicopter.
We asked a ranger at the bottom. Answer: Helicopter. "They just dangle a huge bag of chat and we cut a hole and walk it along the path letting it pour out. Those mountain meadows are easier to do than the steep stuff at the bottom"
I figured "chat" was gravel or something from the context, but had never heard the term before. Maybe I'm the lone idiot, but in case anyone else is confused:
"Chat is a term for fragments of siliceous rock, limestone, and dolomite waste rejected in the lead-zinc milling operations that accompanied lead-zinc mining in the first half of the 20th century. [...] Although poisonous, chat can be used to improve traction on snow-covered roads; as gravel; and as construction aggregate, principally for railroad ballast, highway construction, and concrete production."
In general it just means small crushed gravel. It's usually white or tan. Not sure if it's always a remnant from mining or not. Does a good job on the trail though.
Also egads is everything just mining leftovers?
I would buy one, if $10-16k, and it went 40-50mph and was street legal. I'd ride it to work. Hell I'd ride it everywhere.
Government probably gets better bang for its buck by mounting a pistol onto each of 1,000 flying drones that only cost $300 per unit. That's an army I'd be unenthused about facing.
Except the ones made by their contractors would be $30k apiece for the same thing.
I'm having a hard time seeing why you couldn't take something like a fat bike, and put an electrical assist motor on it, and have troops use that, mostly by pushing it. If you run out of battery, it still works 100% except now you have to push all the weight (or pedal, which could also be used to charge the battery). Gotta leave in a hurry? Use the bike, or just throw it away. Slash the tires and it's pretty much useless for the enemy.
Probably < $1,000 wholesale in off the shelf parts. Def. less than $40mm. Easily carry 100lbs of gear.
Even a bike like that works only on a limited amount of terrain. Add the 100 ish pounds of weight they need to be able to carry in addition to a person and they get bogged down even worse. The ways you can pack them is also really limited because soldiers would need to be able to take everything off and carry it still in the even that it does get bogged down.
> Even a bike like that works only on a limited amount of terrain.
Same as this monstrosity. Except it's $39,999,000 cheaper.
Bicycle infantry units have been, and are now a Thing.
I wonder if the robot actually works better than a power assisted wagon or wheelbarrow, perhaps with caterpillar tracks instead of wheels.
Irony of it being all electric motors/actuators but needing a gas engine to power it so it is very loud.
If it is like the original big dog then it doesn't use electric motors/actuators it uses hydraulics that are driven by the gas engine.[0] Supposedly it gets ~0.5 miles per gallon.
As a submarine veteran, noise is bad.
These will be turned into mounted auto targeting gun platforms if they haven't been already. This is the 2nd time reading about this topic so the DOD PR team is in full spin. This might be a intelligence counter-measure to show the world 'see we dont want to use this tech!' when in reality it has been taken over into black ops world.
I call B.S. A nuclear powered one would probably be pretty quiet.
Are Rat Things[0] really viable?
"MMRTG contains a total of 10.6 pounds (4.8 kilograms) of plutonium dioxide (including Pu-238) that initially provides approximately 2,000 watts of thermal power and 110 watts of electrical power when exposed to deep space environments."[1]
Even if ~110 watts per ~4.8 kilograms is good for legs in agile situations, as opposed to wheels in a slow-and-steady rover, losing one in combat means you've just given the enemy a few kilos of radioactive material.
[0]https://en.wikipedia.org/wiki/Snow_Crash#Rat_Things [1]http://mars.nasa.gov/msl/files/mep/MMRTG_FactSheet_update_10...
It would be exciting to hunt, too. Catch one and never pay your power bill again!
is it range anxiety that drives using a gas motor versus battery pack and solar panels?
It takes more energy per mile for a walker versus a car, so apparently batteries are ruled out so far.
Solar is not an option in practical transportation. The energy yield per surface area is way too low.
So then no AT-AT or AT-ST walkers because of the same problems. Better stick to tanks instead.
Tanks are not exactly solar or battery driven either, am not quite sure what is your point.
They should've gone with hydrogen peroxide based monopropellant linear actuators.