Guidance: A visual sensing platform for robotic applications

Guyue Zhou, Lu Fang, Ketan Tang, Honghui Zhang, Kai Wang, Kang Yang

Abstract: Visual sensing, such as vision based localization, navigation, tracking, are crucial for intelligent robots, which have shown great advantage in many robotic applications. However, the market is still in lack of a powerful visual sensing platform to deal with most of the visual processing tasks. In this paper we introduce a powerful and efficient platform, Guidance, which is composed of one processor and multiple (up to five) stereo sensing units. Basic visual tasks including visual odometry, obstacle avoidance, depth generation, are given as built-in functions. Additionally, with the aid of a well documented SDK, Guidance is extremely flexible for users to develop other applications, such as autonomous navigation, SLAM, tracking.