In automated and traditional manual warehouses, there is a classic tradeoff between the storage density of the system (items per square or cubic foot) and the throughput of that system (items retrieved per unit time). The more dense the storage, the lower the throughput. This research proposes a new storage system design which could provide both very high storage density and very high throughput, using a concept called "virtual aisles." A conceptual model for this system will be developed, including a decentralized control algorithm that manages the simultaneous retrieval of multiple items. A second algorithm will integrate replenishment of items, allowing the system to operate in steady-state. Multiple objectives will be explored, including controlling the system to maximize throughput and controlling it to minimize energy consumed. The second phase of the research explores the allocation of free, or empty, space within the storage system, in a way that minimizes the expected time to retrieve an item. The final phase will develop control algorithms to permit insertion and extraction of items from all four sides, making, in three dimensions, a sort of "living cube," which allows items to pass in and out upon command.

If the proposed control algorithms provide the promised benefits in terms of throughput and storage density, systems that use them could significantly reduce the footprint of internal storage in factories and warehouses of any industry using automated material handling technology. These algorithms would be part of storage systems with decentralized control, which could allow for more flexible and robust material handling in modern manufacturing and distribution. The analytical methods employed in this research (statistical analysis of cellular automata) will also lead to the development of new tools for material handling systems analysis and design.

Project Report

The objective of this project was to investigate the design, control, and performance of a very high density storage system based on a "puzzle architecture" similar to the children’s game, the 15-puzzle. Imagine a 30x40 grid of square conveyor modules, each capable of conveying in the four cardinal directions. Cells in the grid contain hundreds of totes with products to be retrieved and stored in a warehouse, for example. The grid also has a few dozen empty cells to allow totes to "puzzle" their way to the southern boundary of the grid. How should totes be moved within the grid in order to allow requested totes to exit? The first accomplishment of the project was to design decentralized algorithms to control movement of totes in such a system, such that requested totes were retrieved quickly and returning totes found their home locations quickly. The resulting system—called GridStore—was shown to provide excellent performance characteristics, meaning extremely high throughput and very high storage density, which are traditionally competing objectives. A second accomplishment was the development of a modified GridStore system called GridPick, in which a worker or robotic picking vehicle moves along the front edge of the system picking items from totes corresponding to a customer’s order. Totes move to and away from the front face of the grid as required to present the order picker with only the totes he or she needs to see. In this way, the pick face is significantly smaller than in competing, traditional systems and therefore the worker is 20–50% more productive. The third accomplishment of the research was development and testing of a sequencing system for physical items (totes, pallets, etc.) called GridSequence. The system uses the same grid-based architecture and decentralized control rules to allow totes to arrive to the grid in a random order and leave in a specified order. Such a system could be used, for example, to arrange cartons arriving to a robotic pallet building system. GridSequence is the subject of a patent application. Information on all three systems can be found at www.kevingue.com. Broader impacts of the research include multiple academic papers, seminars at several universities, presentations to more than a dozen companies and interested industry groups, a popular website, and YouTube videos. The project also produced a children’s game called Box Rush for mobile devices running the Android operating system. The game illustrates the same principles developed in the research, but in a way that children and high school students can understand.

Project Start
Project End
Budget Start
2009-08-15
Budget End
2013-07-31
Support Year
Fiscal Year
2009
Total Cost
$293,804
Indirect Cost
Name
Auburn University
Department
Type
DUNS #
City
Auburn
State
AL
Country
United States
Zip Code
36849