An endoscope includes a section of an elongated tube for imaging an object by entry in a body cavity. For use with this, a guide assembly includes a shaft sleeve for mounting on the elongated tube by reception thereof. A housing sleeve is supported around the shaft sleeve. Three endless belts are secured to the housing sleeve, for endlessly moving in an axial direction of the elongated tube, and propelling the elongated tube by contacting a cavity wall of the body cavity. A drive sleeve is contained in the housing sleeve, for driving the endless belts. Side projections as side rail portions are disposed on the housing sleeve, for covering lateral side portions of each of the endless belts at least partially, to keep the endless belts slidable. Preferably, the side projections cause curving of a center portion of the endless belts outwards with respect to a transverse direction.