Robots are safer and know how to learn, with advanced vision and sensing systems
Traditionally, industrial robots have been deployed for manufacturing tasks that required brute strength, such as the heavy-payload robots used in the automotive industry, or they were of the speedy pick-and-place variety, the type of robots often deployed in medical or semiconductor applications. In most instances, safety requirements mandated that robots be entirely sealed off in fence-guarded cells to protect human workers from injury.
But the new robots are changing all of that. So-called “safe” robots have begun to be deployed in an array of applications on factory-floor assembly lines, in machine shops, and pharmaceutical and research laboratories. In the past year, newer robots like the UR5 and UR10 collaborative robotic arms from Universal Robotics USA (Stony Brook, NY, and Odense, Denmark) and Rethink Robotics’ (Boston) two-armed Baxter robot (see TinyURL.com/Safe-Robots in the November 2012 issue of Manufacturing Engineering) have been introduced. These robots offer lower-payload capacities and built-in safety capabilities that allow the robots to safely work near humans on assembly lines or in a lab setting. Similarly, ABB Robotics (Auburn Hills, MI, and Sweden) has demonstrated its dual-arm concept robot for some time and is currently testing it at selected customer installations.
Introduced at IMTS 2012, the UR5 and UR10 robots are developed by Universal Robotics, a Danish robotics company that recently began selling these single-arm collaborative robots in North America. “It’s a brand-new classification of robots that’s being adopted by the industry,” said Ed Mullen, Universal Robotics national sales manager. “The global safety specifications that are being developed by ISO are also adopting the word ‘collaborative,’ so based on the written standard, through force sensing and a risk assessment, the use of a collaborative robot unguarded can be implemented.”
The latest ISO 10218-1 and 10218-2 standards for robotic safety is being adopted by the RIA R15.06 ANSI safety standard as well, Mullen added. “With the proper risk assessment, our robot can be implemented in a manufacturing environment and be able to run unguarded because of some of the features that we have implemented in our technology, specifically our force sensing,” he said. “We can actually sense force up and beyond what the robot is using to operate, recognize there is a potential collision and be able to safely stop the robot and not generate a damaging impact force.”
The system does this through a combination of hardware sensor technology and the company’s software algorithms. “It knows the required torque of all the motors. If it sees something up and beyond that value, it recognizes that as a potential collision and will do what we call a Force Stop,” Mullen said. “It’s a combination of looking at encoder positioning in each of the joints, as well as current into the motors of each of the joints, and through our own proprietary algorithms we’re able to make some decisions based on some deltas in each of those and recognize that as a collision.”
The six-axis UR5 and UR10 robots feature relatively light payloads of 11 lb (5 kg) and 22 lb (10 kg) respectively. Aimed primarily at small to medium-sized manufacturers, Universal’s robots are highly flexible, allowing easy movement for redeployment in a factory for new applications, and are very simple to program, with operators teaching the robot the path desired by moving it to multiple points that it learns. To date, the company has installed more than 50 robots in the US, Mullen said.
“Our market is the cadence-driven, repetitive tasks that are being done manually now,” Mullen added, “so it’s for putting products in bins on a conveyor, gluing, cutting, welding, painting, machine tool tending, loading/unloading parts, things like that, where we’ve got really great accuracy specifications and speed specifications, to be able to achieve the throughput with one arm. Our speed specs are 1 m/sec, and repeatability is ±0.004″ [±0.1 mm].”
Mullen said the company’s robots are being used in aerospace, automotive and many other industries. “The ROI component really starts to go through the roof when you add the ability to re-allocate the program, re-teach the program very easily, by anybody in the facility, and to change the location of the automated device at any place in the facility, depending on where the need is,” he said. “BMW and other automakers are using our robots now.”
New 3D Vision-Guided Robotics
Recent advances in 3D vision systems also are moving robotic automation forward, particularly in offering more effective methods for difficult vision tasks such as bin-picking. In bin-picking applications, problems often arise when parts are stacked on top of each other and may be hidden, or interlocking with other parts.
“You have to understand if there’s overlap, if a part is resting on top of another part, and there’s interlocking that can occur,” said Nick Hunt, manager, robot technology, ABB Robotics (Auburn Hills, MI).
“You’re looking at these parts in the bin at different angles. The hardest part is going through the algorithm fast enough to pick out enough targets,” Hunt said. “When you pick one up and set it down somewhere, you’ve disturbed what was there, so the targets that were good targets before may not necessarily be good anymore.”
In January at the Automate show in Chicago, ABB Robotics announced its partnership with vision developer SVIA Industrial Automation (Jönköping, Sweden) to bring SVIA’s PickVision Smart Camera vision system to North America.
Under the partnership, ABB will incorporate SVIA’s technology directly into ABB’s IRC 5 controller in a plug-and-play system, Hunt said. “SVIA is a vision company that has decided to integrate its technology more tightly into our robot controller,” Hunt said. “This is a positive advancement because many vision companies fall short in integrating their system with the robot controller. They have provided users with some icon-based user screens that simplify the programming and operation.”
SVIA specializes in vision systems for machine tool tending with CNC machines. PickVision is highly flexible and easy to use making it a viable technology for a wide range of robotic applications that can be enhanced by giving robots vision capability. The company has more than 600 PickVision installations in Europe, mostly with ABB robots. The system can be programmed on a PC or tablet and it includes remote service capabilities with an iPhone app.
Bin-picking applications have evolved a lot in the past decade, Hunt said. “We did have it 10 years ago, but it was very experimental. Bin-picking still isn’t to the point where it’s one-size-fits-all. It really does need to be customized and tuned for the application. Speed is still an issue, but that’s getting better and better.”
With advanced 3D vision cameras, which now include the sensor technology, sensor-based processing has really improved. “It never ceases to amaze me how accurate Moore’s Law is,” Hunt said. “We’re seeing it all the time—every 18 months or so, the technology literally doubles the availability of sensor-based technology and the speed of the CPU, all at decreasing costs.”
Later this year ABB will unveil a new Cognex integrated vision system embedding the Cognex EasyBuilder interface into ABB’s RobotStudio software, Hunt said, complete with off-line simulation of the camera for diagnosing image and lighting issues.
Vision improvements have steadily evolved to the point where it’s making an impact on new palletizing applications in warehousing, said Earl Wohlrab, Palletizing and Robotics Systems product manager, Intelligrated Inc. (Mason, OH). “Vision for a long time was sort and inspect,” Wohlrab said. “3D vision technology is evolving to a point which makes it more accessible to warehouse, distribution and fulfillment customers. Our tooling compliance is also evolving to keep pace with changes in other technology such as vision.”
In warehouse and distribution environments, the tooling to handle random and unstructured products has to work well with newer, more sophisticated vision systems, he noted. New types of grippers and tooling configurations with force control and other emerging technologies are elevating applications like bin-picking to the next level, Wohlrab added.
Developed initially for spot-welding and assembly applications, the “Gakushu” or Learning Robot technology from Fanuc Robotics America (Rochester Hills, MI) has been used in automotive OEM and Tier suppliers. The patented technology is a software option that is available for multiple robot sizes within the Fanuc lineup, said Rob Totten, Fanuc Robotics applications engineer, including the R-2000, M-900, and R-1000 series robots.
“Gakushu means to learn in Japanese,” Totten said. “We mount a three-axis accelerometer to the end-effector, whether it be the welding electrodes, welding gun, material-handling gripper, or whatever is mounted to the face of the robot. And then the robot basically performs its process repetitively.”
On the first pass, the robot records or memorizes the path, and then from that point it knows how fast it’s going to get, Totten said. “It looks at each individual move to make sure that it can achieve that particular speed. It applies the Learning Technology incrementally to maintain the taught path because one of the major benefits, in addition to speed-up and vibration suppression, is that the path is maintained, requiring a minimal amount of time to set up.
“Traditional methods of speeding up a robot typically include changing your program override speed or modifying acceleration parameters, which have direct impact on the path and the path quality,” Totten said. “So with Learning Vibration Control [LVC] or the learning robot we’re able, by incrementally increasing the speed up of the target program, to maintain the path while at the same time we’re smoothing the path, because we’re actually using the accelerometer data to not only speed up but to improve the path.” The robot goes through 18 or 19 cycles until learning is complete.
The latest mobile robots from Adept Technology Inc. (Pleasanton, CA) combine courier/transporter type robots with a vision-equipped robotic arm for selecting and delivering tools, packages and information within pharmaceutical, medical and laboratory operations. Adept’s Lynx and Mobile Handler autonomous indoor vehicles (AIV) are deployed in medical laboratories and semiconductor manufacturing clean rooms at customers including one of GlobalFoundries’ semiconductor fabrication plants in Singapore.
“Another variant is the Mobile Handler, which is a newer product,” said Rush LaSelle, Adept Technology vice president and general manager of the mobile business unit. “It’s more of a fully autonomous solution in that it has a manipulator on top, a robot that sits on the base with some vision and some sensory input to locate a tool, pick it, and then transport it to another location.
“This is an exciting aspect of the business, because I personally believe that mobile manipulation is really going to usher in a new wave of automation across a whole host of different industries,” LaSelle said. “It’s easy to start in places like healthcare, semiconductor high-value products, but as the technology curve evolves and costs drop, I think you’ll see it become more pervasive.”
Aimed at high-speed pick-and-place applications, the TP80 fast picker robots from Stäubli Corp. (Duncan, SC, and Paris) deliver up to 200 picks per minute for high throughput and short cycle times. The four-axis TP80 fast picker robots typically compete with overhead-mount Delta robots by offering speed and high-precision throughout the entire work envelope.
“The fast picker robots are used in high-speed picking applications for food, consumer goods, pharmaceuticals, photovoltaic and other applications requiring high throughput,” said Chad Henry, Stäubli North American sales manager. “In many cases we’re faster and often more repeatable than the Delta robots,” he said. “These are two very different types of machine kinematics. The Delta machines have three or four long arm extensions so they’re inherently prone to vibration. We have been told that as you get out to the edges of the work envelope, the performance and repeatability can degrade. This is not a problem for the TP80 because of its basic design.”
The fast picker design looks similar to a traditional SCARA robot, he added, and it is designed for much higher speeds, (up to 200 picks a minute) at a lighter payload. “A SCARA robot is in the 80-100 range, and some of the Delta robots are close at about 180 picks per minute,” Henry said.
The TP80 standard features include a maximum payload of 1 kg, very high rigidity, repeatability of ±0.05 mm and IP65 Protection Class when equipped with bellows. The TP80 robots are controlled by the Stäubli’s CS8C controller, which is a standard hardware and software platform across all Stäubli products. ME