Do a googgle on Perendev. He doesn't have a very good reputation. The video is pretty cool. There have been several tries a duplicating his design, but nobody could get it to work. (http://www.fdp.nu/perendev/
I'm very interested in a PMM and have been researching them for about a month now. A lot of the negative talk surrounding PMMs is about how magnets loose their magnetism (guass) over time, especially when they are forced to repel over and over again. I think there is some hype surrounding this. You should be able to size the magnets versus the load appropriately to keep the magnets magnetized.
The best example I have found of a PMM design is by Bedini. It's very simple and looks like it makes sense. He is a big proponent of Resonance Energy (ether, free energy from the active vacuum, etc.). He has created devices based on the theories of Tom Bearden. He has three patents related to Resonance Energy. He actually took his device and broke it up into three patents because he knew the USPTO wouldn't issue a patent for the whole thing (an OU device). He has published designs for his School Girl device on PESWiki.
Anyway, his PMM design is here: http://www.icehouse.net/john34/motor.html
. It's the first drawing. I couldn't figure out the rest of the drawings. I also think it will work because it's only using repelling forces, instead of trying to balance repelling and attracting forces like in the replications of the Perendev device shown on www.fdp.nu
Another VERY interesting design is that of Flynn Research (http://www.flynnresearch.net
). Take a look at all of the drawings and illustrations they have for their PMM and how they generate rotary motion from permanent magnets. I think this is similar to a switched reluctance motor. I'm curious what would happen if you used three magnets in their linear example; does the magnetism increase 8x?