In this paper, dead-reckoning localization errors for automated guided vehicles (AGVs) in indoors quasi-structured environments are studied. Based on error ellipse localization analysis, dead-reckoning errors arising from mismatched AGV wheels radius are investigated. The dynamic model for a differential-drive AGV type with mismatched wheels radius is also derived. The developed rationale shows that sensor fusion is a key feature for minimizing AGV localization errors. Two simulation scenarios and one experiment are presented to show system performance under AGV proportional-derivative (PD)-type control.

Vehicle navigation systems use various sensors and the global positioning system (GPS) to locate a vehicle. This location is then matched to a map database to provide navigation information. Between GPS updates, the vehicle's heading angle and forward speed are used to “dead reckon” its position. Heading angle is often measured by integrating the output of a rate gyroscope. For this measurement to be equal to the vehicle's heading angle, the vehicle should not experience any rotation about its roll or pitch axes. For an automobile, the roll and pitch angles are small and may be neglected for the purposes of navigation. This article demonstrates that this same assumption is not true for a motorcycle. Through simulation, it is shown that for a motorcycle, obtaining a meaningful heading angle from a single angular rate measurement requires accounting for the motorcycle's roll angle. Methods to estimate roll angle and heading angle from available navigation measurements are presented, and two possible sensor configurations are compared. A motorcycle navigation scheme based on these roll angle estimation methods is shown to produce exceptional results in a simulation environment.