Miniature artificial compound eyes for optic-flow-based robotic navigation

摘要

Miniature curved artificial compound eyes (CURVACE) represent a promising tool to assist in navigation of mobile robots. In particular, they feature ideal properties for minidrones to navigate safely in indoor environments due to its panoramic field of view, high temporal resolution, on-board data processing, small size and lightweight. However, the latter limits on-board available resources significantly, including processing power. In this paper, we demonstrate optic flow extraction by means of a CURVACE prototype with on-board processing and using two different methods. First results on fast optic flow extraction with miniature artificial compound eyes are presented. On one hand, 2D optic flow is calculated on-board using a customized version of Lucas-Kanade method at 1000 optic flow vectors per second across a field of view of 180×60 degrees. As well, we demonstrate the on-board processing of 2D optic flow vectors at a rate of more than 21000 vectors per second from the full field of view as well. For this, we utilize an existing assembly language version of the I2A algorithm.