Vision-Based Path Construction and Maintenance for Indoor Guidance of Autonomous Ground Vehicles Based on Collaborative Smart Cameras

Abstract

In this paper, we present a guidance and coordination of autonomous ground vehicles in indoor environment. The solution is based on a set of distributed ceiling-mounted smart cameras with overlapping field-of-view for global coordination. A mean shift based algorithm is implemented to extract a map of the environment. This map is used for a distributed routing of autonomous-guided-vehicles from source to destination. Shortest paths will be calculated and updated in real-time. Our approach fits the requirements of decentralized coordination, real-time environmental changes, as it is the case in production facilities, and portability to other fields of application. First, tests in a test-bench showed satisfying results in terms of reliability, validity and performance.

Decentralised indoor smart camera mapping and hierarchical navigation for autonomous ground vehicles

“…As we are referring to waypoint generation here, it is important to note that we operate on our binary images, , achieved from image preprocessing. For our captured image test, we use the process of [16] to create , while our generated binary environment's subviews are an inherent representation of , themselves. Our captured image, in Figs.…”

“…We utilise mean shift, a technique for the reduction of multimodal feature space into segmented arbitrary clusters, described in [15], to get operable data of camera subviews. We use a histogram to reduce the results of the mean shift algorithm into a binary image, as shown in [16]. A subview S is used to produce an occupancy designation for all points, (our binary image), where and state is equal to one of two values, zero or one, which can be interpreted as occupied or unoccupied, respectively.…”