There occurs in every epoch, a conventional wisdom so entrenched as to have become beyond consideration. A brief history of perpetual motion would look rather static. A ball may roll down a hill, but is unlikely to roll back up to the same height. Wheels rarely turn themselves, when other people are watching, and, despite the evidence of cosmogonists receiving money, one may often find it disappointingly difficult to get something from nothing. Something from something is much easier. Numerous configurations achieve this, but a simple example will suffice. Let us use objects familiar to any child.
I invite you to consider a simple bar magnet. There is an equal force either side. We may label these as we please, since magnets can't hear. Let us call them +1 and -1. They cancel out to 0, which is not much to work with. Let us instead then, just use more of one side of the equation. A ball will roll down a sufficiently steep, infinite hill, until it is eroded. The concept is trivial. All we lack is our infinite downhill. Let us not spend too long looking.
Motion requires space. Let us spread the equation out a bit. Space 2, 3, 5, or 4 if you prefer, but don't bother with too many, magnets in a linear array, poles aligned such that they would snap together. Suspend above, in a manner of your choosing - ball casters in channel, roller track, or similar, a magnet orientated so as to be repelled by face of 1st magnet, attracted to last.
Perhaps unsurprisingly, if your spacing is not too great, it will move from 1st to last. Or, the other way up, the other direction. Starting from rest, it may be configured, by varying height etc, so as to continue after acceleration beyond the final magnet of the array, until it is completely out of field. This is not required, but amusement has its own value. Replace it at the start, and it will do it again, until the magnets get tired. It may be the case that you get bored first.
And here we encounter the problem facing the traditional would-be overuniteer. How to get back to the start. For overunity overall is of course impossible. Manual intervention might be considered cheating. Better would be 'hands-free'.
The arrangement, or similar, cannot be repeated in infinte series, or, at less expense, and more modest ambition, straightforwardly looped round in a circle. This would be to fully encounter the inconvenient opposite force. The forces balance to zero, as we remember from our consideration of a single magnet. And all we have done is, in effect, elongate a single magnet.
We shall now stand back from the shoulders of giants, and see what they overlooked. We shall arrange an asymmetry. We shall just use more of one side of the equation. We shall just use the 'downhills'. Four downhills in a square.
Arrange four arrays in a square. The momentum - after acceleration - is greater than that required to enter the field of the next array at 90^. Avoid too much of the opposite force before the first magnet of each array. Let that beam into the ether by whatever method it deems fit. If you stop in each corner and start from rest again, you are back at the start, only somewhere else. You are back at initial conditions.
As 1 and -1 from nought. We have come full circle. It is written.