Adaptable robot workers for manual production tasks that no one wants to do?
Robot workers hired through staffing agencies?
A radical solution for a brave new world?
Kolektor Digital took on the challenge. Drawing on their in-depth experience of advanced robotics and visual technologies, they needed a 3D machine vision camera with human-like capability. Zivid Two provided the performance required.
Expert in both advanced visual technologies and advanced industrial robotics, Kolektor Digital is a leading provider of solutions for the smart factories in Industry 4.0. Based in Slovenia, the company combines 60 years of industrial manufacturing expertise with cutting-edge digital technologies to create smart factory solutions for the future.
As an early adopter and proactive exponent of machine vision in product quality control systems, Kolektor Digital quickly adapted the technology and applied it to much broader scenarios; in process management, identification, and dimensional measurement applications.
The company has continued to break new ground in visual technologies. Today it is recognized for rapid advances in artificial intelligence, visualization, augmented reality, and virtual reality systems. Its work is often regarded as pioneering and disruptive, in all the most positive of senses.
Being at the forefront of vital industry megatrends, Kolektor Digital became a strong advocate of the democratization of advanced machine vision and robotics. For small and medium-sized businesses and large corporate enterprises, it means lowering the barriers to entry.
In 2018 Kolektor Digital instigated a radical new development program entitled, ‘KoCo Robot Worker’. Its goal would be to produce an adaptable robot capable of detecting its surroundings, manipulating products or trays, activating machines, changing tools, and simplifying visual inspection.
Supported by a team of 50 experts in artificial intelligence and robotics, the KoCo robot worker project laid the foundation for a forward-thinking business model. The purpose; changing the traditional CapEx model in manufacturing to a more agile OpEx model, which lowers the entry barrier for automation. This is achieved by establishing staffing agencies for flexible robots, hiring robot workers to users to help them achieve greater quality, productivity, and reduced labor costs.
At the heart of the KoCo robot worker project lay a real-world need to create an autonomous robot solution that was highly reliable and adaptable in nature. It had to be capable of tackling a multitude of tasks and in a broad range of industrial sectors and applications.
Kolektor Digital customers had long reported increasing difficulty recruiting and retaining sufficient personnel to handle the more monotonous, repetitive, and strenuous production tasks. The KoCo robot worker would therefore address this trend by taking over such tasks.
Robert Bevec, Head of KoCo Development said, “To produce what is an almost universal autonomous robot, able to handle all manner of different low-skilled tasks, flexibility was a paramount design criteria. KoCo would need to be very easy to reconfigure and able to adapt to completely new workplace settings and duties.”
To arrive at such an ideal demanded a core machine vision system capable of delivering significantly elevated levels of data capture, irrespective of the target scene content. It had to identify and inspect objects of almost any material - metal, plastic, or rubber for example, shiny or matt, light or dark.
Both 2D and 3D image-capture had then to be far more detailed and faster than previously experienced. Only then could the necessary data-rich machine-learning be achieved and the derived intelligence shared across an established network of KoCo robot workers.
And while the choice of robot arm and tooling was largely incidental to the project specification, KoCo needed to maximize portability and payload handling. It had to be easy to deploy and handle as wide a range of objects as possible.
Jure Skrabar, Senior Researcher of Optical Metrology at Kolektor Digital said, “Because we wanted to offer customers a fully-integrated, out-of-the-box robot worker, our aim was always to mount the machine vision camera on the robot arm. This of course had implications for the choice of camera. Too heavy and it would limit gripper payload. We had to maximize payload and reach.”
And as with any industrial robot project, maintaining personnel safety in the workplace was a further crucial consideration. The KoCo robot can be configured to a range of applications, from collaborative, cooperative to fenced-in, depending on the required duty cycle. Using its modular approach, the robot can be reconfigured for the required workplace while assuring safety by monitoring the status of safety modules and adapting the speed of operation.
The team developed their system software using ROS (Robot Operating System) and their proprietary KIS4.0 vision platform. With the KoCo robot worker already determined to be robot- and tooling-agnostic, arriving at the final solution would ultimately center on selecting the right machine vision camera for the job.
Kolektor Digital’s long experience of machine vision technologies quickly ruled out a range of alternatives as incapable of meeting the varied technical challenges the KoCo project posed. Very few were capable of delivering the necessary accuracy, speed, weight limit, and safety criteria without compromise.
Skrabar said, “We were actually aware of Zivid through its earliest camera developments back in 2014 and had gone on to use the Zivid One Medium 3D color camera. Its novel approach, through HDR imaging, white light projection, and image quality on random objects, was already a step ahead. And it seemed to support the blue-sky thinking of the KoCo project.”
After the Zivid Two 3D camera launch, Kolektor Digital found the machine vision solution it was looking for. With a form factor of just 169mm x 122mm x 56mm and weighing in at only 940g, the Zivid Two camera served to minimize any adverse impact on robot maneuverability and payload in KoCo’s strictly on-arm applications.
And thoroughly tested against alternative camera technologies in a range of classical machine vision set-ups, the Zivid Two also delivered on image accuracy, precision, and dimensional trueness.
The camera also met the project requirements in terms of image capture speed. Through faster image acquisition and processing, Zivid Two achieved HD 3D image captures in full color in less than 0.3s, some four times faster than comparable 3D scanners.
Further flexibility was assured by the camera’s generous working distance of 300 mm to 1500 mm, and a wide 57º by 35º field of view. And by depending on the principle of white structured light illumination rather than laser projection, the Zivid Two was also inherently safer in collaborative robot settings.
The Zivid Two 3D color camera has become a critical enabling technology in Kolektor Digital’s ongoing development of its ground-breaking KoCo robot worker. Its stringent design criteria of accuracy, speed, weight and safety have been satisfied.
Skrabar said, “We know that customers have struggled to implement vision systems in the past that can reliably handle certain metal, plastic and rubber parts, but with the approach we are taking, this no longer needs to be the case. Even the most complex, poorly lit and ‘shady’ scenes, and the most randomly arranged objects can be resolved and handled efficiently.”
“The integrity of the vision data is paramount of course, 3D image analysis, matching, and machine learning depend on it. Rigorous testing is telling us just how accurate and stable the Zivid point cloud data is. We’re getting exciting performance results, which are now being proven in the field.”
Bevec added, “Remember here that the universal KoCo robot worker is only focusing on implementing those boring, repetitive tasks that customers are telling us that human operators really don’t want to do, or which they struggle to sustain at a high level for a long period of time without tiredness or human error creeping in.” “Nevertheless, whether it’s a low-level machine tending, pick-and-place, or visual inspection task, KoCo has to offer a solution that is not just more reliable and more productive than a human operator, but one which is infinitely adaptable as well. Only then will it make economic sense for the SME user. Happily, our experience using the Zivid Two camera so far is that this will certainly be the case.”
The first KoCo robot workers are expected to enter service in H2 2021.
KOLEKTOR is a multinational industrial group with more than 60 years of tradition in industrial manufacturing and a wide network of more than 30 companies and subsidiaries in Europe, USA and Asia. KOLEKTOR has established itself as one of the leading global players in mobility components and systems, power engineering, as well as civil engineering, water treatment, automation and electrical engineering.
In the age of Industry 4.0, KOLEKTOR's Digital strategic business unit is one of the leading developers of smart factory solutions and system integrators in the CEE region.