Replacing Brawn With Brains: Advanced Software Algorithms Make Automation More Affordable

Produced by Atlanta-based CAMotion, Inc., the PC-based software provides greater intelligence for automated operations, allowing off-the-shelf, low-cost robotic equipment to handle tasks previously requiring more precise and much heavier components. CAMotion, whose name is derived from "computer assisted motion," is a member company of the Advanced Technology Development Center (ATDC) at Georgia Tech.

Though initially targeting manufacturers, the motion control systems could ultimately be used in the service industry, where menial tasks and low wages also make it difficult to maintain a competent staff. Traditional robotic systems have already brought about significant productivity gains in manufacturing processes that involve highly repetitive, difficult, and dangerous tasks. CAMotion, however, is taking aim at processes that are now often not cost-effective to automate: assembly operations, loading and unloading containers or machines, and inspecting products for quality standards.

"Our goal is to help companies apply automation to routine and menial tasks, the very things that companies have trouble finding people to do these days, and the things that often keep U.S. manufacturers from being competitive," said Dr. Steve Dickerson, chairman of the company. "If you look at what people do in factories and in the service sector, direct labor is now almost entirely used for motion. The fact that people have good hand-eye coordination and the intelligence to know the next task means that people in manufacturing are used almost exclusively for moving things around. These tasks do not require the same precision as placing components onto a circuit board, but the large-scale motions across three dimensions have made them difficult to automate in a cost effective way."

CAMotion's first customer is Vulcan Group in Birmingham, Alabama. The company uses two CAMotion systems, one to inspect automobile roof racks after painting, and the other to stack the completed parts in a shipping container. As part of the inspection task, a gantry-type robot moves both the parts and the vision system to acquire 24 different images of each part to evaluate them for paint quality.

"CAMotion created a solution for us when no one else could" said David Boyd, president of Vulcan Group. "We are very pleased with the system; in fact, we believe that it will revolutionize the way painted parts are inspected in the automotive industry."

Three types of algorithms, all developed by researchers in Georgia Tech's Manufacturing Research Center, are included in CAMotion's software: A vibration control algorithm plans the robotic axis trajectory in such a way as to avoid creating unnecessary oscillation. By "damping out" the vibration, the "intelligent trajectory" allows the use of lighter and less expensive components. A learning algorithm helps the equipment to improve its own performance based on feedforward. This typically produces a 10-to-1 reduction in dynamic error. Position estimation software, combining machine vision, encoders, and accelerometers, helps the system know its own location relative to the work and thus how to move to accomplish a task.

"We can build a machine that is not as rigid and precise because our software compensates for that," said Bob Purcell, the company's president. "Therefore, we can build automation equipment with the same level or higher level of performance at less expense. What differentiates us is the ability to produce a machine that is very cost effective for a variety of tasks."

By using standard components and smart software control, Purcell hopes to make automating those tasks competitive given the rising cost and scarcity of direct labor. Beyond the use of standard robotic components, CAMotion will use a single software package for all of its applications.

Changes to numerical tables in the software allow it to handle a wide range of different tasks, providing a degree of flexibility that has been lacking in automated systems. A graphical user interface (GUI) allows these tables to be updated in a very user-friendly manner by three groups: programmers, maintenance staff, and operators. Each group has its own GUI.

"Our one software package fits almost any motion control requirement," said Dickerson. "We have produced software that is very sophisticated yet very generic so that the marginal cost of installation is small regardless of the size, speed, accuracy, or payload of the robotic machine."

Use of standardized parts will not only reduce costs, but will make the equipment more dependable, Dickerson said.

The company is targeting two groups of customers: end users who will put the system to work in their own factories, and systems integrators who may use the company's products as part of a broader solution.

The software behind CAMotion was developed with support from the National Center for Manufacturing Sciences, a consortium of manufacturers. Faculty members working on that project, including Dr. Wayne Book, Dr. Nader Sadegh, and Dr. Dickerson, saw the business potential for the techniques and obtained a $50,000 Faculty Research Commercialization Program grant from the ATDC to move their research toward commercialization.

The company was formed in 1998, and now operates from the ATDC, Georgia Tech's business incubator. The company is Dickerson's second start-up. The first, DVT, Inc., is a leading machine vision company. DVT's product was also based on technology developed at Georgia Tech and is often used in CAMotion's robotics.