Agile Assembly Architecture
Putting together the pieces of the assembly automation puzzle
Mural: Detroit Industry (Detail), Diego Rivera.
Gift of Edsel B. Ford. Photograph © The Detroit Institute of Arts.
Problems and opportunity
In recent years, manufacturing companies have faced enormous difficulties meeting cost, schedule, and quality objectives. New product introductions are increasing at a rapid rate, and global competition has accelerated. Performance requirements for several types of small, high-value, products are becoming increasingly aggressive while the costs of these products must decrease substantially to remain viable in the global marketplace. Examples include magnetic and optical disk drives, laptop and palmtop computers, small medical and automotive devices, and many other kinds of high-density "mechatronic'' equipment.
The parts which make up these kinds of products are getting too small to be effectively handled and placed by humans, and current manufacturing automation is becoming inadequate to do the job at low enough costs. A key part of our manufacturing capability rests with automated assembly, which must be improved dramatically to maintain profitability of today's types of small, high-value products, and to provide an assembly enabling capability for completely new kinds of products in the future.
Current state-of-the-art assembly systems have a number of shortcomings such as costly and time-consuming changeover, parts feeding problems, barely adequate inherent precision, lack of programmability and flexibility in horizontal conveyances, limited robot workspaces, large dedicated cleanroom floor space, line balancing difficulties, and coarse quantization of production capacity.
Meanwhile, however, there has been a tremendous explosion in the power to perform computations and communicate information. Despite these advances, it is generally agreed that this explosion is still in its infancy. In our view, the world of manufacturing has not kept up with this revolution. We believe that this is a most opportune time to bring advanced computing and communication technologies to bear on manufacturing to greatly improve the way we do business. The AAA project targets automated assembly of small, precision products as a good place to start.
Technical and scientific questions
Ours is a new approach to assembly which extensively draws on high-performance embedded computing and the information infrastructure ---an approach which is precise, modular, and extensible in its hardware and software elements---to increase the agility of automated assembly systems. The AAA project is trying to find answers to the following fundamental questions:
- What should be the strategy for allowing distributed computing resources to work together to meet specified assembly goals in the absence of centralized control?
- What are the most effective methods for communication and control in such a distributed system?
- How best can the required functionality be divided among the building blocks?
- To what extent will vendor consultation versus robotic module self-knowledge be needed for design and deployment of assembly systems?
- How best can we program such a system, and how difficult will it turn out to be?
- Can we ensure robust operation with graceful failure modes?
- How does such a distributed approach scale with increasing size and complexity of assemblies?
- How readily can such a system respond to changing requirements?
- What will be the projected cost/benefits for industry?
Our research is proceeding along both architectural and implementation perspectives. An architectural vision which is not instantiated in real, functioning hardware and software cannot hope to address the relevant issues. Our approach is a tightly-coupled one which vertically integrates all elements from the lowest level of sensors and actuators through the higher levels of the manufacturing enterprise.
Our strategic framework called an "Architecture for Agile Assembly'' (AAA) is embodied by new agent-based robotic elements which are precise, modular, and extensible. These elements form sets of (leased or purchased) self-contained software/hardware modules that are capable of being programmed and operated over the Internet, and which can be brought together to form agile "miniature'' factories for assembly. Such factories occupy perhaps one-tenth of the floor space of conventional assembly lines. AAA depends critically on a powerful combination of intelligent networked communication, distributed computing resources using high-performance processors, and distributed sensor/actuator subsystems of novel design.
AAA embodies high-performance computing and communication technologies and methods which allow factories to be quickly designed, set up, and programmed to respond to rapidly changing market conditions. Our goals are to:
- Reduce the time to deploy a new assembly factory from months to weeks, with adjustment to product changes in less than an eight-hour shift.
- Enhance product quality in several ways including reducing parts-workpiece alignment errors to micrometer levels.
- Substantially reduce the floor space required for product assembly.
Relationship between AAA, minifactory, and others
As shown in the Venn diagram above, our Agile Assembly Architecture is an example of what is known generally as the much wider class of an "agile manufacturing system." Since our minifactory is wholly AAA-compliant, it is a proper subset of AAA. On the other hand, there can be other kinds of AAA systems (e.g. an agile machining system) which is also wholly AAA-compliant and shares some attributes of our minfactory.
The AAA/minifactory project began in January 1994 in the newly-formed Microdynamic Systems Laboratory in the Robotics Institute at Carnegie Mellon University, Pittsburgh, Pennsylvania. Our first support from the National Science Foundation began in October, 1995 and continues to the present. Support from NSF grants CDA9503992, DMI9523156, DMI9527190, DMI9900165, and CNS1059765, an initial equipment donation from IBM, a gift from Siemens, a gift from Weber Screwdriving Systems, support from Mitsubishi Materials Corporation, student support from Technical University of Munich and the Swiss Federal Institute of Technology, and contracts from Zyvex Corporation, Penn State University Electro-Optics Center, Sandia National Laboratory, and the Industrial Technology Research Institute of Taiwan are gratefully acknowledged.
- The annotated AAA life cycle
A "cartoon" description of how AAA will be used to design, assemble, program, monitor, and change a complex assembly system.
- AAA principles and philosophy
- AAA interface tool
The users' interface to the system.
- AAA programming
Distributed programming of the AAA agents.
- Automatic calibration
Generating a precise factory model from the actual factory.