... ..can be an interesting practical exercise,if you replace the swinging motion of pendulum (see Milkovic's "Two Stage Mechanical Oscillator " or "Hammer with a pendulum") ...with your intended model to self-rotate.
The simple idea is to use,not only the rotational torque,but also the normal/radial N component of gravity(usually "scattered","lost" in the pivot...),of the oscillating fulcrum/support.
The test,can be provocative,or even more...useful to find a possible next step in our look over the PM subject.
There is no greater failure,except in...no longer our will to try more and more...
All the best! / Alex
....can be your unique "lucky" chance :as you are in a "waiting-room" ,thinking to imagine a gravity self-powered simple machine (a "particular" lever...),in the same easy manner.
All the best! / Alex