A flying robot controlled by a biologically inspired vision system.

摘要

We are developing small indoor autonomous helicopters controlled by biologically inspired vision systems, which can detect the displacement of their attitude with real-time video processing. A helicopter has two CCD video cameras to see landscape objects and has processing circuitry built with an FPGA. A simple vision processing method was implemented and tested. Each image is divided into 800 small areas for edge detection analysis, which are then used to detect displacement. Each of these small areas works as an ommatidium found in the compound eye of an insect. In the context of a typical indoor setting (objects such as desks, walls, and so on) the feedback from visual features were not sufficient enough to realize stable hovering, but the helicopter could stay airborne for a short time. We are now developing a more advanced flying system with 4 propellers, 4 video cameras, and a better processor for the next step, to realize true sustained hovering and free movement.