Some thoughts for alternatives to let you do what you need to; there's a few sections to the post, so its' kind of a lot to read--sorry:
As others have pointed out, http://www.ebikes.ca/tools/simulator.html
is a great tool to learn how different things affect a system under different conditions, and how changing parts of that system (voltage, current limits, which motor is used, etc) fix some problems and create others.
The thing that you need for your project is not power, it is torque. Takes power to make torque *but* the power itself is irrelevant, because you can waste lots of power without making sufficient torque, and that just makes lots of heat and drains your battery pack.
(in trying to create more torque for climbing, you turn up the throttle, which ups the voltage on the motor, which increases current. Usually this would make the motor spin faster, but if you have a motor meant to spin at 20-25MPH because of it's winding, and you're only able to push it to 2-3MPH, most of the power is just going into heating the windings, not moving the motor, because nromally the motor's higher speed that it is wound for would cause a back-voltage (BEMF) to be created, negating most of the voltage being applied, which then negates most of the current being applied, so the motor doesn't overheat. If the motor was wound for the slower speed, then the BEMF would be able to help prevent this while letting the motor operate as needed at the lower speed).
So what you want, if you have to use hubmotors, is ones that have windings specifically done for the very low speeds you are after. Wheelchair and wheelbarrow motors are probably the only ones already made this way, and the only company I know that advertises making those is Goldenmotor. (there are probably others, though). If you go to https://goldenmotor.com/
and click on Hubmotors on the left, then on the right scroll down a little more than halfway, you get to the motors already built to move at those speeds. Unfortunately they are typically for low power as well, so may not work for you. There probably is no existing hubmotor to do exactly what you want.
But you could have existing motors rewound for the speed you are after. You'll want motors with as wide a stator as you can get, at as large a diameter at the magnet ring/stator interface as you can get, because those two things themselves directly contribute to (or take away from at smaller diameters) the torque-making ability of a motor.
Because of your requirement that a single motor be able to support the chair's movements, It might require finding a large-diameter-stator motor, then taking *two* of the motors apart and joining their stators side by side (it's been done before, see Farfle's motor thread) and then winding it to create the necessary speed.
You'll need to work out how much torque you actually need to move the chair up the worst parts of the slope (the sharp rises of climbing over boulder edges, when it also has the worst traction) at the speeds it has to go. I don't know the specific math on that, but it's straightforward.
You'd have to work out what winding you would need to achieve the speed you're after at the voltage you'll be running. If the motors you will be starting with are already in the simulator, or you already have them in hand, you can figure out what that will be by experimenting / in the simulator to see what speed they run at full throttle (for max current and thus torque), then seeing what winding they already have, and using the motor winding math (which I don't know but is in a lot of places on ES and elsewhere, like RCgroups.com) to determine the number of turns of wire you need for the torque and speed you're after. AFAICR you need more turns of wire per stator tooth for lower speed.
Regarding using RC controllers / heli controllers, those are generally designed for free-spinning propeller motor control, and are not suitable for controlling extremely low speed motors under heavy load at that speed. Most likely they will blow up; if they don't they may have trouble controlling the motors at the low RPMs you're needing.
Another way around the problems, as has been suggested, is to use geared motors (preferably non-hub for power-dissipation and size/durability reasons). Use a motor that can spin at high RPM to keep the motor itself smaller for the same power output, then use a high-reduction gearbox for converting that speed into torque. It's not as efficient as a direct drive (DD) hub would be, but it's a lot easier to find stuff that would directly do what you want.
There's geared hubs but these have power dissipation limits; if they get too hot inside, they either kill their hall sensors or damage their plastic gears or both, and if you can get around that eventually the limit is the windings themselves.
The other issue, as pointed out, is most of them have a one-way clutch so they can only spin the wheel in one direction and have no ability to electrically brake.
Or you can use motors that don't mount in the wheel itself, but are separate with their own gearbox and output shaft to drive the wheel (either directly or indirectly).
Most of these don't have a directional limitation, and are designed to drive the wheel both directions, and could also probably electrically brake.
The easiest source for these is old powerchairs (powered wheelchairs), which have a pair of them with axle mounts already on them, for directly driving wheels--often 14" size, but sometimes smaller or larger. Get three sets that are the same so they'll all be setup for the same RPM at the same voltage. I see chairs like this at goodwill and yard sales every few months here in Phoenix, just in my local area, so I'm sure you can locate some in your area. They often go for $100-$200, and mostly all taht's wrong is the SLA batteries are dead, maybe worn out tires/seats/etc. Motors I've seen are always working (but if they're not it's probably either a wire or a brush).
There are a number of types and sizes, and some of htem are over 600-700w rated at 24v--but that's usually a lot more torque at that power and wheel diameter than the DD hubmotors, and that is a *continous* rating (they can take a much higher momentary power, and if you add cooling (like a fan sucking air thru the motor) you could probably double the continous power.
Usually these are 24v systems. Because you don't want to run them fast, you don't want to run them any higher voltage, either. Usually they are meant for around 4-8MPH; you can check the specs for the chair they come off of (or just test it before you take it apart). If you find some that are made for a faster speed than you want, then use smaller wheels/tires on them--they'll run at slower speeds *and* higher torque that way.
Or if you have to use them at a higher voltage, get motors meant to drive larger wheels and then use proportionally smaller ones to get the intended speed out of them at the higher voltage.
They are also usually brushed motors, which means you *can* use one controller per side, and series (or parallel) the motors, simplifying control systems.
If you series them, then you need to use a battery three times the voltage the motors are rated for, so each one will get the right voltage to give it the right speed.
If you parallel them, you need to use a battery the same voltage they are rated for, and a controller rated for at least three times the current you need at maximum, to supply them all with power when they need it at the same time.
The downside is these are generally significantly heavier than the DD hubmotors for the same power, but they are much more suitable for the slow speeds ahd high torque you are after. They're also less efficient than brushless, but again--I think in the right configuration they'll do what you want, though you may have to actively cool them if they're used way above their ratings for the whole climb.
See my old CrazyBike2 thread (and http://electricle.blogspot.com
posts) about these motors, where I abused them at higher power and voltage levels than intended (and did damage one from overheating). At one point I had a chain jam from the frame twisting from the torque these things put out, and before the chain broke, it wrapped around the sprocket and destroyed that, and pulled the rear wheel out of position after destryoing the axle and the rim and spokes.
If I were to build a rock crawling wheelchair, my first iteration of it would use the highest-power (physically largest) powerchair brushed motors w/gearbox that I could find, driving wheels of a diameter that would get me the best speed / torque / object-climbing compromise I could make.
Another thought about the whole system, is that regardless of what kind of motor system, you will want to figure out how much power it's going to take to do the whole climb. How many watts, over how much time.
You can estimate this with a simple slope vs power calculator, of which there should be a bunch on the web (probably are links to some in the calculator threads of ES).
Wild guess (probably wrong), just for the sake of showing the calculation: Say it takes 3000w average over the whole climb, and it takes four hours to make the climb.
That means 3kW * 4h = 12kWh.
That would mean at minimum you'd need a 12kWh battery pack, which is huge.
I have an almost 2.5kWh EIG NMC pack on the trike that weighs around 35lbs+, and is the size of a small stack of books.
Another issue is that the battery has to be able to support the continuous current draw of the whole system, as well as the max peak draw.
Let's say it is 3kW continuous, and 10kW peak, just for some random numbers to calculate with.
If you use a "48v" pack, that makes it 3000W / 48V = 62.5A continous. 10000W / 48V = 208.3A peak.
My EIG pack could handle both of those (it could actually sustain just under the peak draw continously), but not all batteries can do that. Some of them would need to have more pack capacity than will actually be used, just so there are enough cells in parallel to handle the current draw. Most of them probably wont' have an issue with it, but you'll want to size the pack with this in mind, as well as how long the pack needs to run the system for (range).