One of the biggest things that some people have said against the flywheel or other "recharging" ideas is that "its not efficient enough"
the thing that i think you guys are overlooking is that ATs aren't meant for marathon runs (or for long continuous periods of activation either)

the idea is to LENGTHEN the battery life (which also allows for a smaller battery and reduces the weight)
besides, i think you guys are going on the premise that the ATs will be running for upwards of 60-70 percent of the time, only stopping when you lift your foot, but that would be unnecessary. Think of how a car drives, you only rev the motor to accelerate and the rest of the time is more or less coasting. This means that motor activity is a LOT less than what you guys are thinking about.

I think about how running works. as you move your foot forward and it hits the ground, there is a little moment when your foot wants to move forward but friction and traction keeps it in place, THEN you push it backwards to propel yourself forward. (i dont think im explaining this well enough but i hope you see the picture). on wheels, this would translate into something like 5% forward and 95% backwards. we could use that 5% to activate the motor which would strengthen the other 95%.

do you get what im saying? it sounds right in my head but it gets all muddled when i try to explain it.
just use that initial motion caused by the foot (or wheel) hitting the ground to activate a short burst of motor power. you could start it off slowish and accelerate it so that the burst doesnt flip the "stormrider". (it looks like stormriders go longer intervals between steps anyway). this is where computer systems would be in place to figure out what speed to start at and how long it should accelerate for depending on the rider's particular trends of how long they keep one foot on the ground before switching to the other and etc.