As a first step toward an Automatic Flight Control System (AFCS) for Micro-Air Vehicle (MAV) obstacle avoidance, we introduce a vision based autopilot (LORA: Lateral Optic flow Regulation Autopilot), which is able to make a hovercraft automatically follow a wall or centre between the two walls of a corridor. A hovercraft is endowed with natural stabilization in pitch and roll while keeping two translational degrees of freedom (X and Y) and one rotational degree of freedom (yaw). We show the feasibility of an OF regulator that maintains the lateral Optic Flow (OF) on one wall equal to an OF set-point. The OF sensors used are Elementary Motion Detectors (EMDs), whose working was directly inspired by the housefly motion detecting neurons. The properties of these neurons were previously analysed at our laboratory by performing electrophysiological recordings while applying optical microstimuli to single photoreceptor cells of the compound eye. The simulation results show that depending on the OF set-point, the hovercraft either centres along the midline of the corridor or follows one of the two walls, even with local lack of optical texture on one wall, such as caused, for instance, by an open door or a T-junction. All these navigational tasks are performed with one and the same feedback loop, which consists of a lateral OF regulation loop that permits relatively high-speed navigation (1m/s, i.e 3 body lengths per second), with a minimalist visual system (only two EMDs, each EMD uses two pixels). This principle contrasts with the formerly proposed strategy that consists in equalizing the two lateral OFs. The passive visual sensors and the simple processing system are suitable for use with MAVs with an avionic payload of only a few grams. The goal is to achieve MAV automatic guidance or to relieve a remote operator from guiding it in challenging environments such as urban canyons or indoor environments.