I've been thinking about it some more and are you sure you need to move your own weight? I don't think you do. That greatly reduces torque requirements. After all, the whole point isn't to move the person, but to allow the foot that the person is pushing off of to move so that the person doesn't move forward. All you need is enough force to overcome rolling friction so that it's perceived that there is zero friction when pushing off the shoe. I don't know how much this force is but the fact that it's super hard to walk forward in roller blades indicates that much less torque is required. I think that's why the the flex shaft becomes viable. You're not trying to move the person. You're just cancelling out the rolling friction of the shoe. The rest of the rolling force comes from other person's being off-center-of-gravity and pushing on their foot.
Just curious...how did you plan on dealing with the fact that people don't step with their toes pointing straight forward when walking? Or that the foot rotates as you push off of it? It almost seems like a shoe with balls rather than wheels and drive X-Y rollers and a brake would be more natural. You would also need also need multiple forces gauges around the foot so you could know how the person's weight is distributed on each foot so you know which foot is on the ground and which direction the shoe has to move to counteract their motion. Whenever one foot measures zero force you know that the person might be attempting to push off on the other foot. When the deadband is exceeded (when the person is off-center and falling in a certain direction) you move the planted to shoe to counteract that motion. I guess the tricky part is to know how much force differential on the foot indicates that they are off balance since I am able to balance on leg with the weight distributed on different parts of my foot. Also with balls instead of wheels, you can side step.
Maybe you don't even need a motor. Just really good slick wheels and brakes that fire in accordance to the data from the force gauges should mostly keep you in one place. After all, the whole point is make the planted foot slide so in theory if you had a zero friction shoe, you wouldn't need a motor Motors could also be replaced by standing in a slight semicircle bowl so that the slight incline allows gravity to provide the force to overcome the rolling friction. In that case, the main thing would just be to control the breaks so the inertia when sliding doesn't come into play when it's not supposed to.
EDIT: A simpler way to account for people not walking with their foot straightforward might just be to mount the wheels on a plate that the shoe sits on that can passively rotate slightly, possibly weakly spring loaded. If the plate is made to rotate very slightly off center from the foot the same way a chair caster is, it would automatically align itself in the direction the person is pushing. However, the fact you off-center it means that it probably won't work for both walking forward and backwards, since to align properly when walking forward you need to rotate around a point slightly towards the heel and walking backwards would require rotating around a point slightly towards the toe. Also won't work for side stepping though.