Designing Parts Feeders Using Dynamic Simulation Dina R. Berkowitz John Canny dberk@cs.berkeley.edu jfc@cs.berkeley.edu Department of Electrical Engineering and Computer Science University of California, Berkeley, CA 94720 Abstract We consider the problem of designing traditional (e.g. vibratory bowl) feeders for singulating and orienting indus- trial parts. Our ultimate goal is to prototype new designs using analytically- and geometrically-based methods. We have developed a tool for designing industrial parts feeders based on dynamic simulation. Our tool allows us to automatically perform multiple feeder design experiments, and to evaluate their outcomes. These results can then be used to compute the probabilities of a Markov model for the feeder. To demonstrate our technique, we present preliminary results for the design of two simple feeders. Our findings suggest that using dynamic simulation is a promising approach for designing parts feeders. 1 Introduction Vibratory bowl feeders and hopper feeders have prolifer- ated industry as a cost-effective means for reliably orienting parts. These feeders and their transfer conveyors account for nearly one-third of the cost and failure risk of an assembly system [8]. However, the current design of these feeders is a “black art”, based merely on modifications to previous designs and empirical debugging rather than on theory and automated design. Parts are presented to feeders in bulk, and as a result, clustering and entanglement are common. Singulating and orienting parts is a significant problem, due largely to the non-general feeder design solutions that must be developed for the individual parts. The complexity of the parts and the feeder, the number of parts, and the absence of good impact friction models in the literature add to this formidable task. These inflexibilities often result in a 7-12 month turn-around time for each new feeder system, even for feeders that orient the most similar of parts [2]. Our focus in this work is to use dynamic simulation to expedite the design process and to make it more flexible, efficient and robust. We are using Mirtich’s novel near real- time impulse-based dynamic simulator, Impulse [14; 15]. Impulse was expressly developed to simulate parts feeders. It was designed to represent many colliding rigid bodies, and is founded on a new friction-based simulation paradigm Financial support provided by National Science Foundation Grant #FD93-19412 and Advanced Research Projects Agency Grant #B454. Figure 1: A Typical Vibratory Bowl Feeder for Orienting Indus- trial Parts. (Reprinted from Boothroyd [4] p. 32 by courtesy of Marcel Dekker, Inc.) that models impacts more accurately. Impulse’s stable pose predictions have been shown to accurately characterize the dynamics found in industrial tasks such as parts feeding [16]. The primary contribution of this paper is to develop pa- rameter enumeration, analysis, and Markov model-building tools. We have developed a tool that allows us to easily gen- erate and test suites of feeder designs with different charac- teristics and initial conditions. Our tool also evaluates the generated designs automatically. Once a good feeder design has been found, the tool can be used again to simulate a par- ticular feeder over all stable poses of the part. The results of these experiments yield probabilities that allow us to build a Markov model of the feeder. We have used our tool to simulate simple parts feeders modelled after vibratory bowl feeders. We first describe our parts feeding model. We then summarize how our tool is used in practice to drive design experiments enumerated over a user-specified parameter space. Next we discuss how our tool automatically evaluates a design. Finally, we present a block feeder design case study and some preliminary results of a cap feeder simulation. Proc. of the IEEE Int. Conf. on Robotics and Automation (ICRA’96), Minneapolis, Minnesota, April 1996.