| An Automated Guided Vehicle (AGV) is a self-propelled, driverless vehicle generally used to transport materials in a factory or office environment. Until recently, two methods of defining guidepaths have been in common use. Vehicles used in heavy industrial environments generally follow a wire buried in the factory floor. AGVs employed in offices or relatively clean industrial settings often use photofluorescent or reflective materials applied to the floor to define the path network. Both methods offer a safe, reliable alternative to human operators. The paths, however, are expensive to change, particularly in the case of the wire-guided systems.; This thesis describes the development of a navigation system for industrial automated vehicles that will allow them to operate without continuous guidepaths. The navigation system is based on the principle of dead reckoning (DR) with periodic position correction from landmark detection. The dead reckoning calculations are based on very accurate counts of the wheel turns. The landmarks are retroflective paper targets mounted on the ceiling which are located by an on-board video camera. The landmark information is combined with the DR position estimate by a Kalman filter. The control system incorporates a number of adaptive control techniques, including sliding mode control for the steering angle, and parameter estimation for detecting and compensating calibration errors and wear.; A prototype vehicle was constructed which is capable of navigating a path described by a series of commands read from a floppy disk. It has successfully negotiated a variety of paths, and has demonstrated DR random tracking errors of less than one percent of the travel distance. It was concluded that dead reckoning with landmark updating is a viable navigation technique for industrial and commercial automated vehicles. While vision was used for the landmark recognition in this project, a variety of other sensor types are possible, and expansion of the machine's sensor options is encouraged as an extension of the research. |
