Algorithms for automatic guided vehicle navigation and guidance based on linear image array sensor data

Abstract

Algorithms for path planning of an automatic guided vehicle (hereafter referred to as AGV) using data from a Linear Image Array camera (hereafter referred to as LIAC) are outlined in this dissertation. The use of LIAC data for tracking two guidelines on a factory pathway in order to derive a map of the immediate areas around the AGV is shown. The prior forms of path sensing and planning systems for an AGV are briefly described, and compared to the requirements for stable control of a practical system. The system requirements are given, based on control stability, robot performance, and execution time for the algorithms. Based on these requirements, an architecture for a path planner using LIAC data is described with the required algorithms. Conceptually shown as three processor systems (vision, path planning, and motion control), the architecture data rates are shown to match the system requirements for real time guidance of the AGV. The algorithms for the path planning are described, with changes to the existing algorithms given to improve execution speed from 20 minutes to roughly 20 seconds for planning a route. The known obstacles are avoided by this algorithm, with time delayed penalties within the world model. Unknown obstacles are detected from the sensor data and either passed or retreated from based on the estimate of the obstacle region. The information on the area around the AGV is then used to provide continuous data to the motion control processor through the use of spline functions. The motion processor uses the spline functions at a higher sample rate to control the AGV motors and estimate the world position.