OPTIC FLOW AND AUTONOMOUS NAVIGATION

Many animals, especially insects; compute and use optic flow to contro l their motion direction and to avoid obstacles. Recent advances in co mputer vision have shown that an adequate optic flow can be computed f rom image sequences. Therefore studying whether artificial systems, su ch as robots, can use optic flow for similar purposes is of particular interest. Experiments are reviewed that suggest the possible use of o ptic flow for the navigation of a robot moving in indoor and outdoor e nvironments. The optic flow is used to detect and localise obstacles i n indoor scenes, such as corridors, offices, and laboratories. These r outines are based on the computation of a reduced optic flow. The robo t is usually able to avoid large obstacles such as a chair or a person . The avoidance performances of the proposed algorithm critically depe nd on the optomotor reaction of the robot. The optic flow can be used to understand the ego-motion in outdoor scenes, that is, to obtain inf ormation on the absolute velocity of the moving vehicle and to detect the presence of other moving objects. A critical step is the correctio n of the optic flow for shocks and vibrations present during image acq uisition. The results obtained suggest that optic flow can be successf ully used by biological and artifical systems to control their navigat ion. Moreover, both systems require fast and accurate optomotor reacti ons and need to compensate for the instability of the viewed world.