Robots, Cameras and Cables: Avoiding Catastrophic VGR Failures

On April 4, 2014, the robotics industry lost one of its most passionate champions. Adil Shafi will forever be remembered for his vast expertise in vision guided robotics. Generations of students and robotics professionals will continue to benefit from his tireless devotion to sharing his knowledge through articles and numerous webinars. As Adil's generous contributions were integral to this discussion on camera cable management, the RIA Board of Directors and Staff dedicate this article to his legacy.

By themselves, robots lack perspective. Pair them with vision. Now you’re cruising. Welcome to the Big Dance in vision guided robotics (VGR). Attendance is up. More users are embracing VGR’s possibilities.

“Vision allows you to do so much more than just a blind robot can do,” says Phil Baratti, Applications Engineering Manager at Epson Robots in Carson, California. “It opens up opportunities for precision, throughput, manufacturing variations, and product changeovers. There’s so much more flexibility when you have a vision system integrated with a robot.”

Couple sensitive vision technology with multi-jointed manipulators and you’re bound to run into some snags. That’s why the knowledge base assembled by the industry’s forebearers and early adopters is so valuable. We’ve tapped into that vault to prepare a new generation of integrators and users for VGR’s nuances. Even seasoned veterans occasionally stumble on the details.

Do you mount the camera on the robot or at a fixed location? If you put the camera on the robot, which joint do you use? How do you protect the camera and cables from undue wear?

Dollars and throughput are often tied to these decisions. Different robot configurations and their applications call for specific camera placement and cable management considerations. Approach it from the wrong angle and the result can be catastrophic.

Camera Cables No. 1 for Downtime
Often an afterthought, cable management is critical. Corral the industry’s celebrated storytellers and early firefighters. They will tell you decade-old tales when cable failures were common. Even now, cables often spell the difference between success and failure.

In fact, the number one reason for downtime in arm-mount camera applications is cables. It’s not unheard of for a $300,000 cell to go down because of a $300 cable. Fortunately, with new camera standards, robot-mounted cable management has become easier.

“Most of the large OEMs realize the importance of cable management,” says David McCain, Manager Robotic Solutions at LEONI Engineering Products & Services Inc., a cable and dress-out manufacturer in Lake Orion, Michigan. He says it’s usually the Tier 1 and Tier 2 suppliers that could benefit from assistance in cable management.

“Some of them still use what we call the ‘spaghetti system’ with all the cables and hoses bundled together with tie wraps,” says McCain. “When the tie wraps break, the cables catch on the tool or nearby equipment, or they rub on the robot and eventually wear out.”

Either way, the end result is costly. But it’s not always a catastrophic failure. Sometimes the damage can be more insidious. It can be intermittent lost signals between the camera and the robot.

“Camera and vision cables are generally very small in diameter,” says Dave Carruthers, Project Manager North America for REIKU/Drossbach in Trenton, Ontario, Canada. “I put them in the micro category. Every single strand inside there is almost like the hair on your head. When they’re bent and taken beyond their comfort zone, or what we call a dynamic bend radius, that’s when you start to see these strands fatigue and eventually break.”

“Every time one of these strands breaks, there’s less signal going through the wire,” explains Carruthers. “All of the sudden the camera doesn’t get a good signal. A part might not be put in the right place, or the system might get disoriented and totally discharge a perfectly good part.”
 
A cable management system can be adapted for most robot configurations or applications. As shown in the previous photo, a LEONI dresspack was adapted to a 3D laser vision guidance and gauging system.

“Our Robotic AutoGauge Systems require an extremely robust dresspack due to the extensive range of robot articulation required to run a robotic temperature compensation routine,” says Alan Corker, Manager of New Business Development at Perceptron Inc. in Plymouth, Michigan. Corker says the visual fixturing procedure requires highly articulated movements to scan the periphery of the vehicle door and all of its features.

We’ll explore more cable management solutions and camera placement options for articulated, SCARA (Selective Compliance Assembly Robot Arm), delta-style, and even dual-arm robots.

Articulated vs. Fixed Camera
Reputed for their flexibility and might, articulated robots usually steal the show with their twists, turns and rolls. While their dexterity makes this robot configuration one of the most versatile, it raises several important considerations for camera placement and cable management. The more rotations (or axes) you add to the dance, the more opportunities to get tripped up.

When considering a fixed- versus arm-mounted camera on an articulated robot, the application typically determines the camera location.

“The arm-mounted camera is the most flexible in that you can position the camera anywhere in the work envelope,” explains David Bruce, Vision Product Manager at FANUC America Corporation in Rochester Hills, Michigan. “If you have an application with a bunch of conveyors and you use an arm-mount, then you only need one camera. You can move that camera to all the different conveyors and find your parts.”

“The downside of that is cycle time,” says Bruce. “It’s cheaper in terms of hardware because you only need one camera, but you will have to come to a complete stop, take your picture, and then pick the part.”

FANUC’s tightly integrated vision system incorporates software features that address this arm-mounted trade-off.

“Our iRvision® has something called Snap-in-Motion,” explains Bruce. “Because the camera and robot are on the same controller, meaning the same CPU is controlling both, we can actually take a picture while the robot is still moving.”

Bruce says you have to watch your exposure and have the right type of part. “If you need a long exposure time, then you need to go slow enough so there isn’t any blur. If you have your aperture open all the way and you have a lot of light, you can really get away with a relatively low exposure time and move pretty quickly. I’ve tested it up to 1,000 millimeters a second with good results.”

He notes that Snap-in-Motion does not lend itself to working with a conveyor, because typically you have to be moving the camera across static parts when taking the picture.

“The alternative would be fixed-mounted cameras where you have one for each conveyor,” says Bruce. “You’ll have increased hardware costs, but you can run all the vision in the background. So while you’re on one conveyor, you can be firing the camera on the next conveyor. By the time you come to that area, you’ve already found the part, located it, and calculated the offset. All the vision would happen in parallel.”

What about different joint placement? Bruce says the majority of the time the camera is mounted to joint 6 (J6). But some applications call for novel variations.

“For welding applications, they often mount it to joint 5 (J5),” says Bruce. “In the case of a top loader, they might mount it to the carriage. Joint 1 (J1) is a linear axis or RTU (robot transport unit), so it’s still kind of robot mounted, but you don’t have to worry about protecting it.”

In other applications, a camera may be mounted on J5 and a flexible tool on J6. The calibration and tool transformations become interesting, but you gain more flexibility and power. For example, you can apply breakaway tooling on J6, like compliant knives or tool changers, while carrying a camera on J5.

Twist and Shout, No Tie Wraps!
With all this articulated movement, the dance floor gets crowded. If your camera cables are not appropriately dressed, things could go awry. The dress code usually calls for high-flex cables.

“That is a concern with vision cables,” says FANUC’s Bruce. “They are prone to failure, especially in higher articulation applications.”

“Most of our robots have an internal camera cable from J1 to J3,” he says. “But it’s important that the cable from J3 to the camera is dressed properly.”

Bruce says tie wraps can be a big problem. “With vision cables, coaxial cables, you don’t want to use a tie wrap directly on the cable itself. You want to put something in between the tie wrap and the cable, like a grommet of some kind. Typically, when a failure happens, we’ll say ship us the cable. A lot of those times we’ll see the impression where a tie wrap was used.”

He says engineered dress-outs greatly improve the reliability of robot-mounted vision. “There are active dress-outs, where there’s a spring involved, so when the wrist needs to flip around 180 degrees, you’ll get more slack.”

Bruce is referring to the spring retraction system that LEONI and other dress-out suppliers offer.

“The dresspack consists of a spring retraction system and what we call the ‘umbilical’ or corrugated tube that houses all the communication cables and supply hoses running from the robot to the end-of-arm tool,” explains LEONI’s McCain. “Once it’s installed on the robot, the spring retraction allows the end effector to pull the umbilical forward and retract it back into its housing, keeping the corrugated tube and the cables from rubbing against the robot.”

McCain says the idea is to eliminate as much of the torsional stress on the cables as possible. He echoes the precautions about tie wraps.

“You don’t want to put rigid plastic tie wraps on any kinds of cables or hoses, because once the robot starts running, those tie wraps will wear or cut through the cable or hose jacket, causing failure. We use a hook and loop fastener or polyurethane tie strap to bundle the hoses once they leave the retraction system. These won’t cut through the cables or hoses.”

Proper Dress Required
FANUC’s Bruce stresses the importance of protecting the camera cable connectors from the stress of the cable bundle.

REIKU has a patented grommet system called Cablestar just for this purpose. Shaped like a hockey puck, this specialized grommet is made of soft rubber in different diameters depending on the size of the corrugated tubing used to house the cables. Each grommet system has custom, water-jet bored holes, or slots, where you insert the cables it’s designed to secure.

“Once the cables are put inside the Cablestar, they’re held firm, so there’s no more push-pull effect,” says REIKU’s Carruthers. “They’re all isolated in there, so they’re not influenced by the other cables and hoses. The big ones aren’t pulling on the smaller ones. They only have to carry their own weight.”

He says this reduces the strain on connection points and also seals the conduit to help keep out contaminants.

Carruthers recalls a time when camera cables were so sensitive that the automotive OEMs were going through them like cheap shoelaces. They would bundle the camera cables ten at a time and remove each one in succession as it broke.

“Then we put the cables in a corrugated tubing system, and lo and behold, the robot would run for 2 to 3 years. Now everyone recognizes the benefits of cable management.”

He says Drossbach invented the corrugated tubing process in the early 1900s. REIKU uses that process to manufacture tubing for dynamic and static high-tech cable protection. He says they just patented a modular, highly adaptable cable management system called the Drossbach Cable Saver™, which has been in testing at robot and automotive OEMs for the last year with very positive results.

Depalletizing in 3D
Cable management was a critical step for robot integrator Weldon Solutions in a recent depalletizing and machine tending application for a boring mill system.

This video courtesy of FANUC shows the cell in action. The robot uses an overhead 3D area sensor to first locate and pick the parts from the pallet. Then a separate 2D vision system is used to precisely determine the part’s orientation in the gripper before loading it into the boring mill.

In this video demo of FANUC’s latest bin picking technology, you can see this overhead-mounted 3D area sensor in more detail.

“We ran our cables through a corrugated tube,” says Charles Gales, Manager of Automation Sales at Weldon Solutions in York, Pennsylvania. “Because of all the movement of the end-of-arm tool in the machine tending operation, we found that the cables were hanging too low and getting caught up on nearby equipment, so we had to rework that dress-out.”

Gales says experience and proper testing pays off in these situations. He suggests proceeding deliberately and exhaustively with cable management engineering to ensure that you follow the complete path out to those last several joints of the robot.

“If you don’t, your dress-out will be either too tight or too slack,” he says. “Either one of those conditions can create some big problems.”

Cable Management, One Step at a Time
Global automation supplier, Comau Inc. in Southfield, Michigan, takes a modular approach to cable management.

“We generally break the cable up into different sections,” says Mark Anderson, Technical Development Manager at Comau. “We’ll have one section of high-flex cable going from joint 6 down to joint 1. Then we’ll have a low-flex cable from the base of the robot back to our cabinet. If the robot were to be mounted on some type of 7th-axis slide, we would have another high-flex cable that would go through the 7th-axis CAT track and then a low-flex cable from the end of the CAT track back to our cabinet.”

“That way if there’s a failure in one area, you can replace just that section. It’s much quicker,” says Anderson. “For instance, the cable area along the robot is going to see more bending, flexing and twisting. It takes less time to repair that section rather than replace the entire expanse.”

“The low-flex cable is also lower cost, so you don’t have to replace a very long, expensive cable,” he adds.

Even though Comau manufactures its own robots, the company considers itself “a solutions provider” rather than a product provider. This dictates how they approach every aspect of an application and drives innovation. Comau serves the automotive and aerospace industries, among others.

“We review every application with an open mind, independently and objectively,” says Anderson. “We document the requirements for each individual process, such as desired accuracy, expected product variation, system model flexibility whether they’re running one model, two models, or more and the difference between those models, and any specific customer requirements.”

Anderson says by the time they go through this process, it’s usually determined whether a vision guided application is robot or fixed mounted.

“A lot of our customers would rather us mount the camera on an end effector and have a more flexible system. The body shops where we install systems are traditionally designed to accept a very large and wide product variation.” He says these variants can range from sedans to SUVs.

VGR in Six Degrees of Freedom
“The value of our system is that we offer a very flexible vision guided system that provides a six-degree-of-freedom offset to the robot without requiring any calibration,” explains Anderson, who helped Comau develop robot vision guidance technology for drilling holes in the inlet ducts of F-35 fighter jets. “Our RecogniSense® system is very simple to set up.”

“We helped develop the system out of necessity to fill a void in the VGR arena. We’re an integrator at heart. We want to make sure we provide the right tool for our customer.”

“Vision is not a one-size-fits-all tool,” he says. “Every vision system has its own pros and cons. We probably use just as many Cognex systems as we do RecogniSense.”

Anderson says that in addition to the simple setup and not needing calibration, the patent-pending system has a wide area of view with a minimal number of cables and cords. He says they basically have one cord that plugs into the camera.

“We use a process called visual servoing,” explains Anderson. “It’s an iterative process in which you stop, take a picture, make an adjustment, and then take a picture again. Through that iterative process, we drive the robot to the proper position.”

“We’re able to use this visual servoing process and still achieve a very fast cycle time. Image processing happens in a couple hundred milliseconds.”

This video courtesy of Comau shows the RecogniSense technology being used for automotive wheelhouse hemming. This video shows in more detail how the technology works.

Anderson says that the RecogniSense camera is not always robot mounted. Comau uses it in fixed-mount applications for laser brazing. In these cases, visual servoing isn’t necessary because the image acquisition and processing occur in the background.

One camera or two, joints 5 or 6. With articulated robots, the options seem endless. What happens when we add another twist?

Two to Tango
What’s more convoluted than one articulated arm? How about two!

Dual-arm robots, especially those with 7-axis arms, can shimmy into tight work envelopes previously reserved for humans. In machine tending applications, close proximity hand-offs are often required and it may be difficult to get two conventional robots in one area. That’s where these double-handed manipulators are known for their very contained victory dances.

“With a dual-arm robot, both arms have access to the equipment. They can work independently and can hand off parts to each other,” says Greg Morgan, Software Engineering Manager at Yaskawa Motoman in Miamisburg, Ohio. “They can flip a part over or reorient it.”

This video shows a dual-arm robot at work in a machine tending cell.

We first got to know this breed of robot up close and personal in last summer’s collaborative robotics article. Their kinematically redundant manipulators are one of the characteristics that set them apart. Yet dual-arm robots have their own set of considerations, especially when you add vision to the duo.

On-Board Cameras, Internal Cables
Similar to single-arm articulated robots, the camera placement options for dual-arms are varied. Again, the application usually dictates the camera location.

In logistics, where these types of robots are used to pick and place large boxes on conveyors, the camera is typically fixed above the work area. Other applications call for more flexibility in camera placement.

“Having it on the arm is more flexible,” says Morgan. “You can solve multiple stations with a single camera. It’s usually part of the end-of-arm tool.”

He describes one such application where Yaskawa Motoman’s dual-arm robot is used to test equipment in the automotive industry. “We’ve done several systems where we’re checking tire inflation. One arm has a camera and one arm has an end effector. We use the camera to locate the tire’s valve stem. It’s capped, so we have to remove the cap using the end effector, and then test the pressure. Then the robot replaces the cap.”

Morgan says this process is performed pre-vehicle installation. The tires are presented to the robot on a conveyor.

Cameras can also be mounted on each arm. Of course, the application and increased cycle time benefit have to be supported by the added cost of two cameras. Morgan also provides insight on cable management.

“With our dual-arm in particular, we try to route those cables internally. There are already cables and pneumatics running internally to the arms, so we try to run the camera cable through the arm as well. This limits the profile and cable wear.” He says this also applies to the end effector pneumatics and electrical connections.

This video shows a dual-arm robot at work in a biomedical cell.

Vision-equipped dual-arm robots also excel in assembly applications. “Those arms are good for close proximity work to each other and they have a high moment, so they’re good for putting things together,” says Morgan. “Quite often, we’ll use vision to find roughly located items nearby, and that can be done independently with each arm. Then we mate that location information together using our MotoSight 2D software.”

This humanoid dexterity in a small footprint is helping dual-arm robots set their sights on the flourishing electronics industry. Here’s another case where two is better than one, at least in concept.

Two Arms, Two Cameras, Double the Anticipation
ABB’s Dual-Arm Concept Robot (DACR) recently made its North American debut at the ABB Technology Days event in Auburn Hills, Michigan. This 14-axis, dual-arm robot will hit the market sometime in 2015, according to Phil Crowther, Global Product Manager-Small Robots. Until then, attendees were treated to an eyeful when the humanoid concept was put to task in a simulated electronics application.
 
This video shows the DACR’s vision guidance system in action. With a flip of its wrists, the robot uses optional camera-equipped end effectors to transition from locating small components in trays to grasping and connecting them. In the live demo, Crowther moved the trays slightly between tasks to demonstrate how the robot’s vision guidance system easily adjusted to the new location.
 
DACR’s arms and “eyes” can work independently or together. As you can see, all of the cables are internal and out of sight.

Along with Rethink Robotics’ Baxter and Kawada Industries’ Nextage, these double-handed manipulators have other unique features that make them adept as collaborative robots. We will explore those applications in more depth in an upcoming edition this year.

Watchmaker’s Precision
Like their articulated cousins, SCARA robots have several options for camera placement. Cameras can be mounted on one of the joints, or at a fixed location separate from the robot, or both.

With SCARAs the focus is speed and precision. It’s all about cycle time and field of view (FOV).

“Your first question should be, what kind of precision is required for the application?” says Epson’s Baratti. “Based on that, we start building the working distance, the lens type and the lighting.”

“At a minimum, 50 percent of our applications call for a mobile-mount camera,” he says. “There are some real advantages to that. The mobile-mount camera generally allows you to have a smaller FOV, and this is because the working distance is somewhat fixed. But since it’s mounted to a robot, I now have my complete working area available to me.”

“Say you have a five-part assembly,” Baratti explains. “Part 1 is in the left quadrant, so I drive the robot over there and use this small FOV camera to find precisely what I need to pick up. Part 2 is in quadrant one, so I drive the robot to the other side of its coordinate system. The idea that you can now take advantage of the robot’s full coordinate system and still have a small FOV is an advantage of having the camera mounted on the robot.”

As we saw with articulated robots, one of the key differences between a robot-mounted versus a fixed-mounted camera is cycle time. The same holds true for SCARAs. When the camera is fixed, the image processing can take place in parallel with other tasks. The trade-off is the precision afforded by the smaller FOV of a robot-mounted camera.

Sometimes the application calls for a combination of fixed- and robot-mounted cameras.

“The combination of a fixed downward camera and a mobile-mount now gives me the ability to precisely find parts that have been spread out over a very large area,” says Baratti. “First, I find a part in the robot’s work envelope with the large (fixed) FOV camera. Then I drive over to it with a relatively small FOV camera to find the exact pick location.”

What if the application calls for the parts to be precisely placed? Then where do you put the camera? Sometimes three’s the charm.

“There are cases with flexible feeding or bulk singulation applications where you’re required to place the part to 10 microns. That’s a very small tolerance,” explains Baratti. “So we’ll add a third camera which is typically an upward facing camera with an FOV that is just slightly larger than the part you’re inspecting. You can never put all of your faith in the tooling. You can’t assume that it hasn’t shifted when you pick it up with the vacuum or gripper. We use the upward facing camera to build a tool offset, or coordinate system, around that part in the gripper.”

Multiple Joints, Many Options
Also like their larger articulated cousins, SCARA robots can carry cameras on different joints. The application is again the determining factor.

Baratti says that a camera is placed on joint 2 (J2) when the application calls for looking at a fixed plane. “That means you’re not looking at parts at different heights. It’s called a focal plane.”

Now consider joint 3 (J3). “That means I have the ability to move the camera up and down and also rotate it,” he says. “Joints 3 and 4 on a SCARA robot are tied together, so when one moves the other moves. The reason you would mount a camera on J3 is if you’re dealing with parts or fixtures of different height, or at different levels in the Z plane. It’s a popular use when you’re working on a pallet.”

“Having a camera mounted on J2 means that when you work from one side of a pallet to another, the part is going to rotate in the FOV,” explains Baratti. “We have vision tools to account for that effect.”

This video courtesy of Epson Robots shows a wristwatch assembly process with an entire team of SCARAs. At one point in the process, you can see the effect Baratti describes played out on the monitor as a J2-mounted camera moves across a pallet of small components. The monitor shows the camera’s view. Watch as the parts appear to rotate on the screen (about 1:30 into the video).

“That’s what happens when you have a J2-mounted camera going across a pallet,” says Baratti.

“If you have a very complicated application where you’re doing very precise inspections and measurements, you want to keep that part true to the camera. That’s when you would mount it to J4. It allows you to have a better understanding of where a feature is on a part, because the FOV is now presented to the part as it’s rotated repeatedly every time. So there’s an advantage of having a camera actually rotating as you’re moving across the work envelope.”

Baratti says one of the drawbacks of this approach is cable management. “Because now you’re dealing with the camera cable being moved up and down, and rotated, you’re stressing the cable a lot more. So that’s one of the drawbacks of mounting a camera on J3 or J4.”

He says for the most part cable management is not that complicated for SCARA robots. He recommends high-flex cables for any robotic vision application and strongly advises against tie wraps for the same reasons others have cited.

Watch Your Gs
Baratti notes another consideration when working with cameras mounted on SCARAs, especially on joints 3 and 4. Watch your g-forces.

“You’re putting more forces on that camera, on the lens, and all the components when you have it mounted on J3 and J4. So there’s a higher probability of things starting to shift and loosen up in the housing. And in some cases, your iris may even start to change a little bit, or your focus might change.”

“If you know it’s a very aggressive application with high accelerations and decelerations, you’ll want to get a camera lens that has specifications on g-force and how much force it can handle before things start loosening up,” adds Baratti.

When working with a SCARA, Baratti recommends a camera with a small form factor made for robustness.

Dancing in Parallel
Delta-style, or parallel, robots are the quick-step queens. Sometimes called spider robots, their blinding speed is nearly unmatched. When equipped with vision, they can pick and place a variety of items with mind-numbing precision.

One parallel manipulator stands out. Adept Technology’s Quattro™ robot has a patented 4-arm design. Traditional delta robots have 3 arms. The 4-arm delta boasts speeds surpassing 300 cycles per minute. The extra appendage allows it to work with greater accuracy and higher payloads at those rates.

This video shows the Quattro at work packaging boxes of chocolates.

The delta’s tremendous speed and relatively low payload capacity almost always calls for a fixed-mount camera. If it didn’t wreck the camera altogether, an arm-mount would certainly slow the spider’s cycle time to a snail’s pace.

“Since these robots are mounted overhead, it doesn’t leave a lot of room for the camera,” says Yan Banducci, Senior Product Line Manager at Adept Technology Inc. in Pleasanton, California. “The arms would get in the way of the camera’s field of view. Typically, the camera is mounted overhead separate from the robot and positioned upstream from the pick and place locations.”

“Usually the camera is completely separate from the robot,” says Banducci. “Cable management is less of a concern with delta-style or Quattro robots just because of the mere fact that they are mounted overhead. So the cables for the camera, lighting, and robot typically go up and over the structure holding the robot.”

“It’s not a very complex cabling job,” he says. “The structure is already in place. Usually there’s a very large metal frame holding the robot. The easiest way to get the cables out of there is to run them along that frame.”

As One
“For Adept’s products, our controls are actually integrated into the Quattro robot,” says Banducci. “So the brains of the robot are inside that robot base. There’s no separate cabinet for the controls and amplifiers, which would require extra cabling.”

“You want to make sure that you have a rigid mounting for the camera,” explains Banducci. “Typically, we recommend that the camera be mounted to the same frame as the robot itself. You always want to make sure that the camera is in the same position relative to your robot.”

“So if the whole frame were to get bumped, like from a forklift or some other piece of equipment, the camera and robot would move in unison. There would be no disturbance of the relative position of the camera versus the position of the robot.” Banducci says this is important for calibration.
 
“One of the advantages of our system is the fact that our vision is tightly integrated with our controls,” says Banducci. “Think of a person. Your hands and your eyes are both controlled by the same brain, so it’s easy to pick things up and move them around. We like to think of our robot the same way.”

Indeed, the eyes of a robot are the windows to a world of possibilities. So while March Madness is over, our Big Dance is just beginning.

Check out this article for more Dos, Don’ts and Applications in vision guided robotics.

RIA Members featured in this article:
ABB Robotics
Adept Technology, Inc.
Comau Inc.
Epson Robots
FANUC America Corporation
LEONI Engineering Products & Services Inc.
REIKU/Drossbach
Weldon Solutions
Yaskawa Motoman