Visual homing: Surfing on the epipoles

We introduce a novel method for visual homing. Using this method a robot ca n be sent to desired positions and orientations in 3D space specified by si ngle images taken from these positions. Our method is based on recovering t he epipolar geometry relating the current image taken by the robot and the target image. Using the epipolar geometry, most of the parameters which spe cify the differences in position and orientation of the camera between the two images are recovered. However, since not all of the parameters can be r ecovered from two images, we have developed specific methods to bypass thes e missing parameters and resolve the ambiguities that exist. We present two homing algorithms for two standard projection models, weak and full perspe ctive. Our method determines the path of the robot on-line, the starting position of the robot is relatively not constrained, and a 3D model of the environme nt is not required. The method is almost entirely memoryless, in the sense that at every step the path to the target position is determined independen tly of the previous path taken by the robot. Because of this property the r obot may be able, while moving toward the target, to perform auxiliary task s or to avoid obstacles, without this impairing its ability to eventually r each the target position. We have performed simulations and real experiment s which demonstrate the robustness of the method and that the algorithms al ways converge to the target pose.