A distributed vision infrastructure for multi-robot localization

Abstract

Localization is one of the critical issues in the field of multi-robot navigation. With an accurate estimate of the robot pose, robots will be able to navigate in their environment autonomously with the aid of flexible path planning. In this paper, the infrastructure of a Distributed Vision System (DVS) for multi-robot localization is presented. The main difference between traditional DVSs and the proposed one is that multiple overhead cameras can simultaneously localize a network of robots. The proposed infrastructure is comprised of a Base Process and Coordinate Transform Process. The Base Process receives images from various cameras mounted in the environment and then utilizes this information to localize multiple robots. Coordinate Transform Process is designed to transform from Image Reference Plane to world coordinate system. ID tags are used to locate each robot within the overhead image and camera intrinsic and extrinsic parameters are used to estimate a global pose for each robot. The presented infrastructure was recently implemented by a network of small robot platforms with several overhead cameras mounted in the environment. The results show that the proposed infrastructure could simultaneously localize multiple robots in a global world coordinate system with localization errors within 0.1 meters. Copyright