As I figure [and have consulted with some physicists], there no reason why this would not work: object orbiting a planet in one direction, then, half way trough, hits something that propels it in the other direction [at orbital speed, of course], orbits the planet the other way around and hits something that propels it in the other direction again and then the cycle continues back and forward, keeping the object in orbit, but not in a full orbit, but, in this case, just orbiting, essentially, one side of the planet. Now, that (if) this works, there should be no reason why it has to be the whole half of a planet - it can be any distance at the orbits circumference. Even 100 meters. Right? Now the most interesting part. If we put this object a, say, evacuated tube with magnetic system to propel the object back and forward, provided that we deal with power issued and losing-orbit-while-accelerated problem[*] and more power to counter the "dead-weight" of the whole system[**] (magnetic fields actually would push the tube/ system of the object and stay on top of it) - would it [the tube/ system] just levitate relatively stationary somewhere around the path of what would be an orbit? [*] could this be dealt with just making the system a disk, "donut" or something like that while cutting down power consumption at the same time? [**] if the system is at high orbit where friction is no a problem and the whole system is separated in part, there would be, essentially, 3 parts - the object and two [maybe even one is enough] part that actually orbits the planet in opposite direction waiting to bounce at the object and there would be no "dead weight" . In a .. flying saucer scenario it is a bit more complicated as forces on the system accelerating the object somewhat cancels out (although there would be a time delay.. ).