Automated Guidance of Vehicles Using Vision and Projective Invariant Marking
Research output: Contribution to journal › Article
Control based on image information offers great possibilities for advanced automation. Here, a flexible system for study of visual servoing has been developed by extending a standard computer with commercially available instrumentation. A robot moving on the floor is controlled by image feedback. The conclusion is that an industrially useful behavior of automated guided vehicles (AGV) in flexible warehouses could be obtained using available technology and the principles presented here. A flexible system increases the requirements on good interaction with an operator. A man-robot interface has been designed, based on an operator manipulating interactive color graphics overlaid on images of the working scene of the robot. The interaction is done specifying obstacles and stations rather than explicit path coordinates. Visual recognition in a three-dimensional scene has to cope with the problem that objects have infinitely many poses, which may not all be stored. A key contribution in this work is the design of marking symbols, which are described by invariants under parallel projection. The invariants are based on integrated measures and are thus insensitive to individual pixel errors. The constructed symbols can be used for robot marking or as signposts. The marking problem formulation is new and so are the results and the analysis.
|Research areas and keywords||
Subject classification (UKÄ) – MANDATORY
|Publication status||Published - 1988|