A biologically-inspired and resource-efficient vision system using mobile mini-robots for obstacle avoidance

Abstract

This paper describes a resource-efficient vision system for the mini-robot “Khepera”. It is implemented to perform a fundamental approach of robotic navigation, which is obstacle detection by using an optical flow algorithm. This is inspired by visual perception of insects. The optical flow field is evaluated by implementing the sum of absolute differences (SAD) operation of block matching while the mobile mini-robot moves on texture plane. The SAD is realized in an additional processing module based on a Field Programmable Gate Array (FPGA) solution which can perform various tasks in parallel; for example, image processing, camera control and symbolic protocol transmission to the mini-robot. As a result, the Khepera equipped with 2D camera and the additional module is able to autonomously move avoiding collisions with obstacles, called ego-motion. The 30 × 16 SAD operation is processed by computing 16 pairs of pixels in each operation, which accomplishes the real-time approach. The proposed implementation reduces processing time and power consumption.