|
|
|
|
   
|
||||||
|
|
INSIGHT Vision-guide robotics means a single still image from a camera on the robot's end effector can be used to adjust the robot's path in 3- D space. Material-handling issues in manufacturing continue to have significant bottom-line impact. Consider the high cost of manufacturing defects. They drive up production costs dramatically, costs that are eventually passed on to the consumer. A part that does not fit properly due to a manufacturing defect often requires custom repair work, or can force the manufacturing engineering team to discard a partially assembled product. In the automotive industry alone, this quickly translates into hundreds of millions of lost dollars. The convergence of innovations in science and engineering, particularly robotics, robust sensors, vision imaging and guidance technology, and computing horsepower translates into an opportunity for an emerging science we call vision-guided robotics (VGR). This new field could generate two billion dollars in parts-feeding applications alone, based on the following assumptions: * Vision will be the sensor system of choice in the future, increasing from 2% of current robotic sensors to more than 50% by 2005. * There are nearly 1 million industrial robots in the world today, and 80,000 new robots are being installed annually across a number of manufacturing industries: automotive, aerospace, consumer goods, electronics, and pharmaceuticals. * Studies indicate mobile robotics are projected for 3500% growth in units, from $665 million US to more than $17 billion by 2005. * Mean time between failure for robots has increased significantly, driving down core costs. In 1999, robot prices were 40% lower than in 1990, while labor compensation in manufacturing rose by more than 30% during the same time. * Last year, North American vehicle production was approximately 12.2 million vehicles. In 1999, the global market for automotive parts was estimated at $550 billion. Nearly every carmaker has been pressuring suppliers for annual cuts averaging 3% to keep automotive prices flat. Such trends show no signs of abating. Five technologies form the basis of VGR: robotics, vision imaging science, robust sensors and lighting, computing power, and vision guidance technology. Robots have been used in manufacturing plants for decades, with the automotive industry being one of the first to place them on factory assembly lines. Japan, in particular, welcomed robots in industry to compensate for its postwar labor shortage. Intended to replace humans who performed simple, repetitive tasks, robots provided obvious benefits: robots were efficient, could work in environments unsuitable for people, and were capable of working at 100% uptime. The first generation of robots was blind, however, making them suitable only for repetitive tasks so organized that fixtures and programming could allow them to achieve the same results consistently, i.e., drilling holes, spot welding, spray painting, or assembling circuit boards. Advances in robotic technology in the areas of weight capacity, precision, tolerances, work envelope, and speed have provided greater degrees of freedom and enhanced manufacturing quality and productivity. As a result, the range of robotic applications is expanding. Again, Japan, through a project titled the Robotic Challenge in the 21 st Century, is aggressively pursuing changes to laws that currently prohibit robots from operating in such facilities as hospitals and nursing homes. Machine vision and vision science has been used in manufacturing since the 1980s, and is used to manufacture every automobile in North America. According to the Automated Imaging Association (AIA), the world machine vision market of $6.2 billion in 2000 addressed just 10% of potential applications. Growth figures in this area are forecast at 20% per year through 2005, and robots are expected to represent a large percentage of the overall growth for future vision science applications. For 2000, North America experienced a 26.4% increase in the number of vision units sold, and a 25.8% increase in revenues. Vision market growth in Europe also is forecast to grow 20% per year through 2005. A quarter of the world's global vision system suppliers and nearly half (47.1%) of Europe's machine vision companies are located in Germany. Vision technology comprises many approaches and involves three main tasks: image acquisition and formulation, image processing, and image analysis leading to a decision-making action. In image acquisition, the use of digital cameras is on the rise, with great improvements being made in resolution and granularity. Mega-pixel cameras (2000 x 2000 is common) are starting to enter the market. Flexible telecentric rather than fixed-focal-- length lenses are finding increased use, in addition to piezoelectric motion devices useful for providing precise motion to the camera body. Lighting technology, specifically LEDs, represent a giant leap forward in terms of providing uniform, compact, customizable, and long-lasting illumination for machine vision applications. The synergy between reliable video technology and LED lighting is what we refer to as the robotic eye concept. The robotic eye integrates both video and lighting in a hardened, heretically sealed, compact industrial package that can easily fit on the robot arm. Ten years ago, industry could not integrate video and lighting in such a reliable yet compact configuration. With Computing power, the many
advances in computational processing and communications bandwidth are
vital ingredients of vision-guided robotics. For example, the introduction
of the Vision-guidance technology covers the ability of a system to locate objects in 3-D space. The VGR system must be capable of arriving at the position of an object in X, Y, and Z axes at the precise point of roll, pitch, and yaw to be able to guide the robotic arm to handle its operations. Without this guidance, the robot is blind. Four technologies are in use here: single-camera 3-D (SC3D), stereo vision, laser triangulation, and visual servoing. The first, SC3D, is based on the use of a single conventional CCD video camera that takes advantage of 3-D geometry and principles of perspective distortion to calculate the 3-D position of parts using a single still image from the CCD camera mounted on the robot end-effector. The robot controller then uses this information to adjust its path. The major advantages are that this technology exploits a low-cost conventional sensor/computing platform, it's easy to calibrate, and cycle times are short. At present, SC3D is intended for use with rigid parts with stable dimensions, and is not suitable for such flexible parts as plastic blow-molded products whose dimensions vary significantly. Stereo vision contributes the ability to calculate the depth or distance of features or landmarks of a given object relative to the sensor, i.e., to construct a depth map. Algorithms underlying stereo vision have been around for decades, but challenges remain in successfully identifying and locating corresponding object features in camera images. Our company is currently involved in a research project with the National Research Council of Canada's Vision and Sensing Group that addresses the challenges of focusing this technology on industrial part-handling applications in an unstructured bin environment. The major advantages of this technology are its independence from part dimensions and the maturity of the underlying principles. Laser-triangulation aspects of vision guidance for manufacturing use the same principle as laser-targeting systems for military aircraft to acquire air or ground targets. By painting a surface with a laser beam, a laser-- triangulation sensor determines the depth and orientation of the observed surface. Typically packaged in an integrated, hardened casing with a CCD camera and low-power laser diode, the more sophisticated sensors have scanning lasers that project a plane instead of la spot onto an object's surface. This laser plane projection and its degree of distortion can be analyzed to obtain orientation information. Such laser-triangulation technology is suitable for VGR applications involving grinding, edge tracking for welding, waterjet cutting, and coordinate measuring and verification of dimensions. Using laser planes also means the sensor is impervious to even extreme ambient lighting changes, as with weld arcs. Lights, sensors, lasers, and software are part of the vision- guided robotics concept. Visual servoing is a term that refers to real-time visual feedback control of a robotic manipulator. Today, the majority of vision-guidance systems for industrial robots are based on the "look- and-- move" concept, where a camera snaps an image of the target object and the robot's control analyz\es and reports a position for the robot to achieve. This method can be ideal for a wide array of fixtured-part applications, as long as there is absolutely no possibility of the part moving between the "look" and "move" functions. Accuracy may be lower, too, as the calculated part position is directly related to the accuracy of the "hand/eye" calibration (the offline calibration to relate camera space to robot space). If calibration was wrong, so would be the calculation of part position. Visual servoing as a concept is independent of hand/eye calibration since it has the ability to deal with changes in real time. The computationally intensive algorithms required, however, kept it strictly in the domain of the research and development laboratory for a long time. But the continuing availability of inexpensive computing power, and expanding high-bandwidth Ethernet communications links to most major industrial robots, based on an open component object model software architecture, makes visual servoing a realistic application for manufacturing today. We believe this technology will revolutionize the way robots see and interact with their applications. Component-based software is the heart of the effort behind building and supporting robust vision-- guided robotic solutions. Just as VGR represents a convergence of robotics, computing, and vision sciences, the supporting software architecture, which we call the eVisionFactory (eVF) brings together a number of math and statistics-- based elements. These include advanced image capture and analysis, object location and pattern recognition, defect detection and analysis, PLC communication, and network-enabled Internet support. It's the standardized componentbased software that makes it possible for this theory of marrying image processing and robotics to solve manufacturing problems in the real world. To avoid re-inventing the wheel, our company originally set out to source this software environment. The ideal product would allow our engineers to package our VGR algorithms, robot-communication routines, and third-party applications as reusable components to ensure continuous enrichment. It would allow us to reorganize and reconfigure these components for various manufacturing tasks, and support these solutions in the field. New Technologies Converge for Change in Vision-Guided Robotics No single development environment on the market met these requirements. Most companies developing software environments for vision avoided involvement in final application development. Such lack of depth in functionality was sub-optimal for our VGR needs. Our efforts also showed a number of vision-development efforts relying on proprietary processing cards or cameras. Such a pull- through effect, (using software to sell hardware) is a major flaw, since proprietary compoents almost always compare unfavorably in price or ormance with standard platforms such as Wintel. Basing our software development efforts on Microsoft's component object model (COM) technology pays a number of benefits. Components are reusable, which not only makes the developer's job easiier, it also lowers cost and quickens turnaround time. COM components also expose clear interfaces to other components, providing access only to necessary information. This makes the resulting software more stable, as there is less chance for data corruption. As for the future of VGR, manufacturers are already benefiting from this technology for improved inspection, part-handling, and assembly functions, and the technology can be employed for machining operations, spray painting, and identification of parts. Leading robotics manufacturers such as ABB, Fanuc, Motoman, and Nachi, say as much as 50% of all new robot installations will feature some form of intelligent vision. Synergizing the related technologies of vision, robotics, software, and Internet support will mean dramatic growth for industry. Want More Information? Automation will be part of the conference offerings at this year's IMTS Manufacturing Conference, sponsored by SME, Sept. 4 - 11 in Chicago. For more information, call Customer Service at (800) 733- 4SME or check out SME's web site at www.sme.org. For more information on vision-guided robotics from Braintech, contact SME. Applications VGR, SC3D, and eVF are not just technologies that exist in R&D, but are realizing real-world manufacturing gains. In February, Braintech announced that it received an order for two eVF developer systems from Marubeni Corp. (Tokyo, Japan). These systems will allow Marubeni application engineers to develop vision-guided robotic applications initially for the Japanese automotive industry. The solutions built will be sold and supported by Marubeni under a separate runtime license agreement. A total of 19 VGR systems are to have been installed through the first quarter of this year at a major automobile manufacturer's assembly plants through an order from ABB Inc. Canada. General Motors also has a VGR system in place at its Oshawa, Ontario plant as part of its inspection system for automotive air conditioners. Being able to provide an automated process that can see and react to inspection, part-handling, and assembly tasks will replace fixed mechanical systems that are hard to calibrate. TI Automotive Group of North America uses VGR technology for the plastic welding and drilling operations involved with producing blow- - molded automotive fuel tanks. Combined with laser technology and robotics, the VGR system provides TI robots with the ability to make adjustments when welding and drilling each part. "Plastic is affected by variables such as temperature and humidity, so an accurate, fully integrated vision system to guide the robot is critical to this process," says Barry Mitchell, general manager, ABB Flexible Automation.
Did you find this material interesting? Do you want more information of this type? Comment via FEEDBACK
Source: Manufacturing Engieering
Please patronize our many sponsors, affiliates and advertisers today so that we may bring you more advanced services tomorrow. Have you seen the great deals from top brand name manufacturers?
|
|
|
