It was postulated using a wheel (disc type) rotating at a low RPM (say 60 RPMs) with several gyroscopes mounted on bearings around the edge but on the flat surface, with the gyro's flywheels at a 90 degree angle to the plane of the first wheel.
With the gyros all running at high speed, when the disc/wheel is rotated the gyros would remain in position relative to each other and not impede the rotation of the wheel/disc.
If each gyro was locked in place through 90 to 180 degrees of the rotation of the wheel/disc and then released for the rest of the wheel/disc's rotation one could generate a semi-linear thrust.
Keep in mind that there would NEED to be a pair of wheel/discs...