Sizing Up Robot Controllers
Recently, a special ceremony was held at the Smithsonian Institution in Washington, DC, to recognize the contributions of an assembly pioneer. An industrial robot affectionately known as "Alice" was the guest of honor. General Motors Corp. (Detroit) donated the robot to the National Museum of American History.
Unlike her hydraulic controlled predecessors, Alice was the world?s first commercially significant, computer-controlled electric robot. When the robot debuted in 1978 in assembly applications at Rochester Products, a GM division that eventually became Delphi Corp., it featured another robotic first—a special programming language that allowed it to be controlled off line.
Today?s generation of industrial robots trace their roots to the machine now enshrined in the Smithsonian. In fact, most robots share common physical features with it. But, many similarities stop when it comes to the controller.
New technology, such as open software interfaces, PC-based controls, high-speed digital servo networks and smart sensors, is transforming the world of robotic controls. Benefits include reduced floor space, simplified programming, greater reliability and increased performance. As the trend continues, robots and their controllers will ultimately have a new form factor that will be less expensive, easier to deploy and more flexible.
"The basic control architecture of the robot industry is changing," notes Joe Campbell, vice president of marketing at Adept Technology Inc. (San Jose, CA). "It?s a fundamental change that will shrink controllers to a fraction of their current footprint, replace hundreds of individual cable conductors, expansion slots and vision cables with a digital servo network, and distribute power electronics and servo processing into the very body of the robot."
Since the transition away from early hydraulic drives and mechanical timer-based controls, industrial robots have consisted of three major subsystems: controller, power amplifiers and mechanism. Controllers traditionally did the motion control number crunching and executed the user programs. Power amplifiers took the signals from the controller and regulated the power to the motors on the mechanism.
"Controllers and power amps used to be housed in a single enclosure as big as a washing machine or small refrigerator just a few years ago," says Campbell. "Wiring between the components was considerable. Signal cables connected the power amps to the controller, and fat power and signal cables connected the power amps to the mechanism."
Now, controllers are often offered as separate rack-mounted components. In fact, Campbell claims that all but the lowliest controllers today offer some level of scalability, typically with an array of slots for optional circuit boards. Simple digital I/O is offered on low-end systems. High-end systems offer vision, additional axes of motion control, force sensing, and third-party boards for network connectivity and functionality.
To provide greater ease of use, robot vendors are incorporating PC and networking technology into control systems. These interfaces reduce the need for advanced computer skills to program the robot, making it a much easier tool to use.
"Although software has always been considered to be more progressive, given its flexibility, hardware is now also playing a major role in the newest architectures," says Jan Bosteels, product manager for Advanced Motion Controls (Camarillo, CA). "The use of high-speed networking technology and optical fiber links, as well as the use of the latest DSP (digital signal processing) technology have introduced significant leaps in system performance. Even servomotor technology has been pushing the technological envelope."
According to Bosteels, end users are looking for versatility, expandability, and optimal price and performance ratios in robot controls.
PC-Based ControlDemand for improved human-machine interfaces, increased flexibility and lower operating costs is driving an ongoing trend toward open, PC-based controls. "Robot controllers that leverage the widely available PC hardware and Windows operating system are dramatically changing the landscape of the industry," notes Dick Slansky, a senior analyst at ARC Advisory Group (Dedham, MA). "PC-based open control software has been established in industrial automation as a flexible and reliable control solution."
Until recently, robot users were forced to use a single proprietary motion control technology. While single solutions offer some benefits, most manufacturers do not like the constraints of relying on only one vendor.
"Proprietary controller architectures are inflexible and offer inferior development environments," claims Dave Faulkner, executive vice president of marketing at Cimetrix Inc. (Salt Lake City). "They have higher life cycle costs, are difficult to interface to third-party equipment and require special training for nonstandard programming languages and operating systems. The end result is that proprietary systems take too long to design and build, and allow little reuse.
"PC-based motion control provides a cost effective solution and allows for flexibility, better performance and more programming choices than proprietary controllers," claims Faulker. He says the growing use of PC-based systems is having a profound effect on industrial motion control applications. For instance, it is:
- Lowering overall system costs. Motion controllers that can be plugged into a PC lower the overall system cost because they are often based on off-the-shelf components and help to reduce development time with easy-to-use software. Upgrading them can often be as simple as taking the old controller board out and installing a new version in its place.
- Making motion control programming easier. Using standard programming languages available on the PC, such as C++ or Visual Basic, makes developing motion systems software easier because the developers don?t have to learn proprietary languages and can implement other parts. More and more engineers are comfortable with the PC platform.
- Making motion easier to integrate with other systems. By using the PC as a control platform, motion system designers can easily add other features, such as data acquisition and machine vision. Using standard application development environments, they can program the motion, vision and data acquisition in the same programs and have a central location for all of their system control and monitoring tasks.
- Making motion systems more flexible. The PC is giving users more flexibility in designing systems. With all of the hardware and software available for controlling, monitoring and analysis, the PC is a very flexible platform. PC-based systems are much easier to modify to meet both present and future needs.
Today, manufacturers are demanding lower costs, higher performance and increased reliability from robots. But, the core components of most robot systems—encoders, motors, connectors, cables, bearings and castings—don?t offer any big wins in the cost reduction department. And the basic architectures include some very expensive elements, such as fat multiconductor cables. To address those issues, many robot vendors are turning to open architecture systems and high-speed digital networking.
Open NetworkingIn an open architecture motion control system, all the specifications are public. Any vendor or user can design add-on products for it. The cost of equipment development is reduced and components can be quickly redesigned to meet market needs. In addition, end users can modify equipment to meet specific needs or add technology developed in-house to gain a competitive advantage.
When robot builders and systems integrators use a common platform to tie subsystems together, Faulkner says end users "can easily upgrade subsystems to increase performance, leverage today?s open PC architectures, and take advantage of software reuse to shorten project schedules and improve product reliability."
"An open robotics system requires independent selection of the mechanics, motion control, vision, application development environment, operating system and hardware platform," adds ARC?s Slansky. "An open system approach relies upon ?componentizing? each element of the solution, resulting in an integrated system that is faster, less expensive and more flexible than a proprietary controller."
The concept of networking started out as a means to connect terminals, and then PCs and intelligent devices on the factory floor back to a host computer. "The early applications were limited to data being delivered up, new programs delivered down and MIS happily overseeing all of it," says Adept?s Campbell.
"We then saw a rapid growth of PLC networks, including peer-to-peer communications, that got faster and faster. And finally, the device networks evolved, essentially replacing hard wiring for I/O and signal interfaces, including data rich applications like simple motion control."
Analog vs. DigitalTraditionally, motion controllers used servo drives with a ?10-volt analog signal. This simple control system allowed end users to integrate drives and controllers from different vendors. But, the analog interface had numerous disadvantages, such as susceptibility to noise and bulky, expensive cabling.
"The traditional architecture can be described as ?centralized,? with all signals running from the source to a central computer or controls rack," says Edison Hudson, president of MetaControl Technologies Inc. (Morrisville, NC). "Typically, motion control is handled by one specialized controller or board, machine vision by another board or standalone rack, while digital and analog I/O functions add additional hardware.
"Communication between these various subsystems is often through traditional serial communication channels, or by a backplane bus, such as VME," adds Hudson. "Cabling of these systems represents a major constraint on performance, and more important, limits reliability, as hundreds of conductors are required to route signals to the central control chassis. Overall, the traditional approach is cumbersome, physically large and invariably results in a solution that costs $10,000 to $30,000, depending on performance."
Analog controls are being replaced by digital servo networks, which offer numerous benefits, such as:
- Reduced installation costs. Large multiconductor cables are difficult to install, often requiring a dedicated wire way or large-diameter conduit. Distributed servo network communications dramatically reduce the wiring requirements to a few twisted pairs.
- Reduced footprint. The distributed control architecture eliminates the servo amp panel and the motion interface board and expansion slots in the primary controller. That results in a 70 percent reduction in panel space. Fewer cabling and less wiring provides a cleaner, less expensive system. Cimetrix?s Faulkner claims that cabling accounts for 10 percent to 15 percent of the cost of a traditional controller.
- Improved reliability. Cables have historically been a reliability problem, as each wire crimp and contact connection point is a potential failure point. The distributed servo network architecture replaces hundreds of contact points with a simple, standard network connection.
- Increased performance. Distributing the servo processing to the individual DSPs reduces the load on the main controller CPU, leaving more capacity for other functions, such as machine vision or conveyor tracking.
- Simplified scalability. Additional functionality is added via a simple network connection, rather than worrying about how many open slots and how much power is available from the main controller chassis.
- Simplified troubleshooting. Network modules can easily be isolated for troubleshooting, and replacement is reduced to a few simple power and network connections.
"This processing scheme delivers the best performance and flexibility, but also demands a highly deterministic time-based network," claims Campbell.
In traditional robot architecture, there are heavy multiconductor cables and connections between the controller?s motion control board and the power amps, and between the power amps and the robot mechanism. "These cables and the traditional motion control interface board can be replaced by a network and its twisted pair physical layer," says Campbell.
In addition, the digital servo network also becomes the means to add controller features. In a traditional architecture, the controller included expansion slots and their attendant power supply capacity, connectors, rack space and increased footprint. Now, additional features such as digital I/O, general purpose motion control and additional mechanisms can be added by simply connecting the module directly to the network.
Digital BenefitsAccording to Bosteels, benefits of using digital servo network technology include simplified wiring, more robustness and better diagnostics. "The migration to a completely digital architecture, down to the servo drive level, allows for much better diagnostics," he points out. "By having access to internal drive and feedback sensor data via the digital servo network, remote diagnostics and preventive maintenance become possible. That is a very important aspect for an expensive piece of machinery that can cause important losses during downtime."
"High-speed digital servo networks are enabling radically new control architectures for industrial robots," adds Campbell. "Given the clear benefits to users and manufacturers alike, more robot manufacturers will adopt distributed servo network architectures in the future."
Today, end users can choose from several different open interfaces and networking protocols, such as Ethernet, FireWire, SERCOS and SynqNet. Each technology has pros and cons that must be carefully considered.
Ethernet was developed more than 20 years ago as a high-speed serial data transfer network. The most common Ethernet data transfer rate is 10 million bits per second (Mbps), although many companies are migrating toward fast Ethernet, which features a data transfer rate of 100 Mbps. The next generation of Ethernet is capable of 1 gigabit per second (1,000 Mbps). Ethernet is generally incorporated into a motion control system through a standalone controller connected to the PC or network via a standard Ethernet cable.
FireWire was developed in 1986 and is named for its speed of operation. The IEEE-1394 multimedia connection enables simple, low-cost, high-bandwidth real time data interfacing. New motion control architectures built on standard FireWire networking technology offer the ability to effectively distribute the computing required for motion control applications between the PC?s main processor and DSPs in each servo drive. FireWire, a trademark of Apple Computer Inc. (Cupertino, CA), is also known as i.Link, a trademark of Sony Corp. (Tokyo).
SERCOS is an acronym for "serial real-time communications systems." It is standardized as IEC 61491. The German Machine Tool Builders Association initiated the development of SERCOS as an open digital interface in 1986. SERCOS is an open controller-to-intelligent digital drive interface specification designed for high-speed serial communication of standardized closed-loop data in real-time over a noise-immune, fiber optic cable. It has been criticized as being too slow for robotic applications.
SynqNet is an all-digital motion control interface for connections between controllers and drives. The physical layer of SynqNet is based on IEEE-802.3 standards and 100Base-TX, the physical layer of Ethernet. The data link and application layers of SynqNet are specifically designed for motion control applications. It replaces the noise-prone analog interface between drive and motion controller with a real-time digital network connection. Benefits include high performance, ease of use, flexibility, low costs and reliability.
"If anything, the PC is making the task of choosing the best control and motion solution more difficult than ever," laments John Walewander, marketing manager for the Compumotor division of Parker Hannifin Corp. (Rohnert Park, CA).
"Open architecture systems are great for users, but less attractive for manufacturers," adds Advanced Motion Controls? Bosteels. "That will continue to be an eternal dilemma. Also, open architectures tend to have more layers that reduce overall performance. As system performance demand increases, the use of open architectures becomes more difficult."