You may have seen on the news lately that some countries, including Germany, flew in seasonal farm workers to harvest their fruits and vegetables, despite pandemic-related travel restrictions. This left many of us wondering: We have come so far with technological advancement; why do robots that can pick our crops not exist already?
Alas, the act of picking, a relatively elementary task for humans, poses a great challenge for robots. In manufacturing, this manifests itself as the infamous bin-picking problem. The monotonous task of unloading a container one part at a time and sorting parts according to type may not seem complicated, but it is a problem so notorious that its solution has even been referred to as “the Holy Grail” of robotics.
The main reason is that robots can neither feel nor see in the same way humans can. Each instance in bin-picking is unique, as individual objects differ in terms of size, shape, texture, etc. Before it can perform picking, a robot has to carefully assess the piece and learn how to grasp it correctly. This process is repeated every time it encounters a new item. Unsurprisingly, achieving sufficient speed is a challenge, and there is a high potential for error. Without robots to carry out such tasks, valuable staff resources remain tied up, and productivity suffers.
Fortunately, people have been researching technological solutions for some time now, and there are many good offerings on the market. Here are three innovative solutions to the bin-picking problem, which are currently being offered by start-ups.
1. Universal gripping with granule-filled gripper and vacuum
Traditionally, gripper solutions have been based on the human hand, meaning they have tended to be multi-fingered. The large number of individual joints to control and the need for force-sensing creates complexity in terms of both software and hardware. Therefore, researchers have looked into alternative approaches, such as the use of a single mass of granular material instead of a hand-like gripper. This works because granular materials, such as sand, are able to transition between a soft or "unjammed" state to a hard or "jammed" state, facilitating universal gripping.
The jamming transition is utilised by FORMHAND, a young start-up and spin-off of the Technical University of Braunschweig, in their innovative technology. The unique design of their universal gripper consists of a granule-filled gripping cushion, combined with an electrically-generated vacuum. When the vacuum is applied, the granules become compact, and the gripper can grasp the target item. When the airflow is turned off, the gripper returns to its loose state, and the item is gently released. The beanbag-like design automatically adapts to the shape of complex objects, automating bin-picking tasks and freeing up employees for other, less monotonous work.
2. 3D vision with stereo sensors
Just as many bin-picking solutions have tried to recreate the human hand, they have also focused on recreating human vision. In order for robots to perform bin-picking tasks effectively, they need to be able to detect unfamiliar objects with precision and to distinguish between similar objects. This is where 3D machine vision comes into play.
Founded in 2015 by former employees of the Institute of Robotics and Mechatronics at DLR, Munich-based start-up Roboception’s mission is to provide affordable 3D vision solutions. Their powerful stereo sensors allow robots to perceive their environment in real-time and accomplish a wide variety of tasks. The combination of accuracy and speed is achieved by the Semi-Global Matching (SGM) method. Unlike some market solutions, the interface is extremely user-friendly, requiring little prior knowledge of robotics. Sensor capabilities can be further enhanced when combined with the rc_reason software suite, which offers customisation for specific applications such as bin-picking.
3. Fast learning with AI-powered software
Another problem with random bin-picking is that robots require lengthy training in order to be able to grasp unfamiliar objects in an efficient manner. Advances in Machine Learning are allowing scalability and automation of the learning process. In other words, robots can learn much quicker than through conventional explicit-programming methods and perform better in unstructured environments.
San Francisco-based start-up Osaro produces AI software for automating the bin-picking process. With their software, Machine Learning facilitates rapid learning, making it ideal for high-velocity inventories. This AI-supported learning environment, combined with a superior vision system, enables accurate detection of even the most challenging objects. Gentle placement is also guaranteed, and robots can make quasi-independent decisions, with minimal supervision required from humans.
If technological solutions like those mentioned above are implemented, automated and error-free bin-picking can quickly become a reality on your shop floor. The associated benefits are numerous, from increased cycle times and improved OEE to improved workplace safety and more time for employees to complete other, less tedious tasks.
Still not sure how the latest bin-picking technologies or other Industrie 4.0 innovations could increase productivity in your production environment? If so, feel free to send us a message at firstname.lastname@example.org with details of your problem. Drawing on our industry expertise and AI-enhanced search methods, we conduct extensive research and suggest the best technology solutions tailored to your specific situation.