Industrial robots and positioning equipment traditionally perform their tasks by following programmed, fixed paths along x-y-z coordinates. Some of them augment and modify these programs with inputs from contact, proximity and vision sensors to accommodate changes and variations in the nature and position of the work. But what about operations in which proximity doesn't provide enough precision, or vision can't see what needs to be done?
"The classic example is putting a peg into a hole," says Dave Gavel, technical expert, robotics, at Ford Motor, Livonia, Mich. "Using position control, any misalignment will cause the robot to jam. If you're assembling the sun gear into a planetary set and you don't know the positions of everything, or if the positional uncertainty due to stacking tolerances exceeds the clearances, you have to do a blind search."
Much like we humans grope for a light switch in the dark, robots can use today's touch technologies to do that blind search and gently align heavy components into precise assemblies such as gearboxes. "Touch sensing for us entails the ability to sense surfaces and sometimes textures on manufacturing work pieces, allowing for smarter and more-precise material handling," says Henry Loos, Jr., application engineer, Applied Robotics, Glenville, N.Y.
Touch also can enable machines to handle delicate items without crushing them, reliably pick up soft or varying objects, and work in close proximity to people with little risk of injury. As part of the never-ending quest for the perfect android, emerging technologies promise to give machines human-like ability to assess and accommodate variations in the hardness, texture and temperature of touched materials.
Ford Relies on Force Control
Today's state-of-the-art touch application is "typically applied with a six-degrees-of-freedom, force-torque sensor that's mounted on the tool flange or part of a tool changer and connected to the robot controller through an analog/digital control box," says Nick Hunt, manager of technology and support, ABB Robotics North America. In the U.S., ABB standardized on force-torque converters from ATI, Hunt says. "There are other ways to get the six inputs, but that's the typical way."
See also: The essentials of presence sensors
ABB recently introduced Integrated Force Control, a consolidation of discrete software features that previously were available only in the company's machining or assembly Force Control offerings. The module makes it possible to automate complex tasks that previously required skilled personnel and advanced fixed automation, such as machining and small parts assembly that require dexterous handling of work pieces and tools. Hunt says it can dramatically reduce the stress robot programmers are under when faced with processing parts of complex and varying geometries.
"The essence of force control, whether you buy it or roll your own using our core, is the ability for the robot to feel its way around," Hunt says. "It enables the robot to adapt to the application, not the other way around."
For example, in a casting deburring operation (Figure 1), the robot typically grinds or mills a rough casting to a specific contour. "If it's hard-wired, when the robot gets to a big burr, it grinds too hard," Hunt says. "It's hard on the robot, the tool, the grinder motor and the part."
Force control allows the user to easily program the contour. You can walk the robot through the operation by hand — set a point, move the robot, set a point — to teach the robot the contour of the part. Then you can put it in auto, test and run. "Now when the robot encounters a burr, it slows down —you specify the force," Hunt adds. "If it's a big burr, you can set it to grind in several passes. Imagine trying to program that as a straight robot program. That's a mighty complex program."
On a torque converter assembly line at Ford, "We had ergonometric issues with people doing three vertical installs — a hub splined into the turbine, a one-way clutch into the torque converter and the impeller into the pump — because they are blind searches, the parts are heavy, and the parts have high inertia," Gravel says. "The impeller was the hardest because the mating features are 180º apart." Using force control, Gravel adds, an assembly robot can apply a small force and oscillating motion. When the part goes down, the robot knows it's in position, and it can move on.
"We want minimal interaction with side walls so we don't damage delicate seals and surfaces," Gravel says. "We have to stay on-center and gentle. But the torque converter is a sheet-metal assembly, and our grip is not always perfectly centered. So we set the horizontal force command to zero — the robot will run away from contact. Now we can teach the robot by pushing it with a finger. You can't program this, but you can do it with force control."
By using an external force-torque sensor, a tool frame with three force sensors and three torque sensors, and relating the readings back to the base frame, Gravel says you always know which way to move the robot. "We can essentially operate the robot as a virtual, spring-mass-damper system and control forces, oscillation periods and magnitudes."
Tactile Sensors Add Sensitivity
"Force control issues always have challenged robot grippers," Loos says. "In the past, a gripper was either open or closed, with the same amount of force always applied to the work piece. With the advent of servo-based grippers, modulating the grip force became possible. Now touch-sensing technology allows us to complete the loop and directly sense the surfaces of the work piece and the forces applied to it."
Applied Robotics is developing this capability using piezoelectric sensing. "The advantage to us has been the ability to develop grippers that can tightly control the force applied to delicate work pieces of varying sizes and shapes without having to pre-program a series of opening and force profiles," Loos says. "The main disadvantage is that in the case of intelligent force sensing, we need more processing power and a little more time to carry out a grip operation." He says the power available in modern processors more than makes up for that limitation.
"The sense of touch provides information about the forces between a machine and an object," says Yaroslav Tenzer, PhD, postdoctoral research fellow, Biorobotics laboratory at Harvard School of Engineering and Applied Sciences, and cofounder of TakkTile. He says existing tactile technology is difficult to customize, fragile and very expensive, which limited its use outside of research labs, but technology now is emerging that is robust enough for factory use.
"For example, it is possible to identify the moment when a robotic gripper touches an object, so that the gripper can be properly aligned to an object during bin-picking," Tenzer explains. "Then it is possible to identify the position of the contact and correct the motion of the gripper, if required. Finally, touch sensing makes it possible to measure and maintain grip force on an object during manipulation."
Tenzer and TakkTile cofounder Leif Jentoft, a PhD candidate in the Harvard Biorobotics laboratory, are commercially developing touch sensors that use a tiny air-pressure-sensitive digital barometer of a type already commonly used in cell phones and GPS units. Vacuum-sealed in elastomeric compound and mounted on gripper (Figure 2), the sensor detects how much pressure the robot places on grasped objects. The elastomer provides traction, conveys pressure from the surface to the sensor and protects the sensor, which can tolerate up to 25 lb (11 kg) of direct pressure and withstand impacts from hammers and baseball bats. In operation, the sensor can detect forces as low as one gram.
"Current robots require extensive setup, but industrial machines are becoming smarter," Jentoft says. "This means that setup time can be reduced while increasing the functionality of the robot. For example, while closing the gripper on an object, the system could identify when a firm grasp is achieved and automatically verify the thickness."
Shorter setup time also helps using robots for short-run productions in applications such as injection molding. "Industrial robots, such as the ones from Rethink or Universal Robotics, would be the first ones to benefit from tactile sensing technology," he says. "By sensing the force required to operate or compress a product, the technology could also prove useful for characterizing products such as switches, buttons and baked goods."
Widespread adoption of tactile technology will require not only further development of sensors, but also of software and controls that can use sensor data. TakkTile uses digital interfaces that simplify integration with industrial controllers.
When developing a touch application, it's important to consider a number of factors, including the range of forces and sensitivity, space limitations and the manufacturing process.
Also read: Fundamentals of collaborative robot safety
"TakkTile sensors are developed from off-the-shelf components, and therefore, they're easily customizable to required shape and form," Tenzer says. "They're also robust and extremely sensitive." The sensors are being used now to better control gripping forces when installing piston rings on an engine assembly line. "This prevents smashing parts when any are misaligned," he adds.
Giving Robots the Human Touch
Research continues around the world to develop technologies that promise to give robots human-like ability to sense temperature, pressure and humidity by contact. One example, described in a paper from Technion-Israel Institute of Technology in Haifa and reported by LiveScience.com, uses gold nanoparticles covered with organic connector molecules called ligands integrated into the surface of a plastic commonly used to make water bottles.
When the plastic is stressed, the distances between nanoparticles change, altering the electrical resistance. Temperature and humidity also affect the distance between the nanoparticles. "By using a combination of software and hardware operations, it is possible to isolate the values for humidity, temperature and touch — making the sensor three-in-one," reports study author Hossam Haick, a professor of chemical engineering and nanotechnology at the Institute. Haick says that compared with other designs, the device is about 10 times closer to how real human skin senses the environment.
Altering the thickness and material of the plastic surface allowed researchers to control the sensitivity of the sensor. Changing the properties of the plastic "allows measuring a large range of loads, ranging from tens of milligrams to tens of grams," Haick reported. The study is detailed in the June 2013 issue of the journal Applied Materials & Interfaces.
Like their superhuman strength, endurance and precision, the ability to sense and control motion based on properties and forces at the machine/tool interface expands the potential applications of robotic, material-handling and assembly equipment. It also allows the robots to be programmed easily and to work more safely alongside the dwindling number of factory workers they haven't — yet — replaced.
More from Control Design: