Manufacturing assembly systems use manipulation devices to separate, position, and orient small parts. These devices are a critical component of today's flexible and rapidly reconfigurable assembly lines yet the automated design of manipulation plans is often more an art than a science. This dissertation presents an automated approach to the design and evaluation of vibratory bowl feeders, conveyor belt feeders, and mobile robot manipulation plans This research is an empirical approach to the design of manipulation plans. For each device: (1) experimental data is collected, (2) that data is represented as action operators, (3) the actions are sequenced into accurate and efficient plans using a genetic algorithm, and finally (4) the best plans are analyzed and evaluated for that particular manipulation device The primary contribution of this dissertation is the coupling of empirical data, action operators, and a genetic algorithm based planner to produce good designs for vibratory bowl feeders, conveyor belt feeders, and mobile robot manipulation devices
