Gripper Disruption: Topping Off a Robot’s Arm

Grippers gain high-tech friends
Advanced chips, on-device computing, 3D vision, haptics, new materials and designs, predictive data, analytics, machine learning, and miniaturization are beginning to co-locate at the business end of a robot’s arm. And best of all, this evolution of the robot gripper into a higher order tool for both fabrication and feedback, makes a lot of sense.

“Robot grippers and end effectors are having a profound impact on industrial machines,” says Mike Bacidore, editor in chief of Control Design in his piece: Grippers and end effectors pave the way for the future of robots. “Grippers have increased the versatility of robotic arms exponentially. There is seemingly no end to what an effector or gripper can do. From pressure-sensitive handling and assembly to 3D-vision-based detection and recognition systems, end effectors can be equipped with practically any needed device.”

For cobots, the way forward is pretty clear. If a gripper can take on the task of being aware and careful of human collaborators working nearby, why not make it capable of other jobs in addition to safety?

The logic is compelling. A gripper is already intimate with nearly every workpiece fabricated and/or assembled on an automated production line. In Factory 4.0, it’ll most likely be the sole handler of everything. As such, it will be at the epicenter of an unprecedented flow of valuable data, much of it created at the intersection of gripper and workpiece. It would make sense to enable a gripper to harvest and communicate that data…and to learn from that data. 

In a connected factory, the gripper could create, harvest and communicate data with other machines, either upstream or downstream, as well as with remote grippers working jobs from raw materials to finished products. And, when Cloud connected, it can communicate with any other gripper anywhere on Earth…or anywhere else in the cosmos.

For example, one initial and obvious job for an enlightened gripper would be quality control or quality assurance. According to MetricStream “quality underperformance, i.e. quality defects [cost] between 10-30% of a company’s revenue”, plus lots of customer dissatisfaction. Saving a bit of that 10-30% should be motivation enough to smarten up a gripper’s work life.

Piling on ever more jobs for a gripper to do produces ever more data to be analyzed for improvements in performance, process and products for every aspect of fabrication…and for logistics as well.

Developers who have recently moved intelligent innovation to the end of a robot’s arm are being rewarded for their foresight. Grippers are now doing things they’ve never done before—and doing them well. So well, in fact, that investors are pouring millions of dollars into advancing gripper tech even further.

Opening gambit
Early on, the nascent smart gripper is being used to replace previously human-only functions. Repetitious, mind-numbing but necessary jobs for humans like machine tending, welding, bin picking, component assembly, and quality control are now quickly succumbing to grippers infused with a bit of intelligence.

RightHand Robotics has developed a gripper with fingers and a suction cup. It uses its 3D computer vision, embedded sensors and machine-learning algorithms to see and feel what it’s doing, helping it to adapt to new objects and situations. Investors have ponied up over $11 million for the promising gripper. As one convinced investor, Mark Valdez, from Playground Global, put it: “It’s not enough to simply build a good gripper, you need the machine intelligence to connect what you can see with what you can do.”

Soft Robotics is another with inventive gripper tech coupled to machine learning. It just picked up $20 million in investment for its “soft” three-fingered gripper system. The interesting idea here was that Soft Robotics built intelligence into the gripper material itself so that the robotic system could easily reach out, grasp and manipulate objects.

Google’s Sergey Levine and team are training hand-eye coordination via continuous feedback. The large-scale robotic grasping project—and 800,000 grasps—has grippers learning. The gripper observes its own motions and corrects in real time. “All of these behaviors,” says Levine, “emerged naturally from learning, rather than being programmed into the system.”

Andrew Ng’s $175 million AI fund has sprouted Landing.AI with its visual inspection system and learning algorithm that, “after being trained on just a few images, can be used to spot imperfections in small electronic components or camera lenses.” Foxconn’s assembly robots will use the vision system to improve upon Apple smartphone production.

Perfecting and scaling smart gripper tech will have a major impact on productivity and hasten the dominance of smart manufacturing, which is forecast to “increase production capacity up to 20% while lowering material consumption rates by 4%.”