Soft Robotics introduces a low-cost, AI-driven picking system for its rubbery grippers

Massachusetts-based Soft Robotics has already shown us some pretty impressive capabilities for its rubbery robotic grippers, but the demos have mostly been straightforward picks off of a conveyor belt. The startup’s latest trick is the much sought-after activity of bin picking.

It’s a surprisingly tough task, given the complexity of picking up even several of the same object arranged at different angles. Most solutions require a fair amount of offline training and modeling in order to exhaust all the potential scenarios the arm might encounter when retrieving something from a bin.

The new SuperPick system, on the other hand, leans heavily on the company’s malleable, air-filled grippers, which are able to conform to and pick up a wide array of different shapes.

“It is designed to autonomously retrieve, sort and fulfill orders, with little to no human intervention,” Soft Robotics CEO Carl Vause told TechCrunch. “We do no training. We don’t do the offline learning, we don’t create the 3D models, because all of the things that are driving you to create those models, we can solve with the material science. We have people throw stuff in front of the robot and it can pick it up.”

The system is capable of more than 600 picks per hour, according to Soft Robotics. Like most robotic systems, SuperPick is designed to replace dull, repetitive warehouse tasks that often lead to a high turnover rate — and honestly, warehouse tasks don’t get much more monotonous than picking, sorting and placing objects from bins.

The system is available now. No word on pricing — that’s largely dependent on the job for which it’s being configured, but the company promises it will be available for a fraction of the cost of existing systems, courtesy, in part, to off-the-shelf components like webcams.