KUKA Brings First Intelligent Bin Picking System to Market

KUKA AG offers the new KUKA.SmartBinPicking software. This latest system offers the perfect combination of artificial-intelligence (AI)-based recognition of objects and precise path planning of the robot.

Mainly, KUKA.SmartBinPicking software and associated components equip KUKA robots to bin-pick specific parts. For the first time, this automation solution is available as a vision toolkit package, making it easier to integrate into production.

Easy Pickings with the Bin Picking Application

It’s easy for a human to reach into a bin and pick up a screw without hitting an arm on the bin edge. A robot needs precise motion planning to reach into the bin at the correct angle and grip the right part. With proper preparation, the robot can relieve humans of tedious, monotonous bin-picking work, and do so reliably, quickly and without tiring.

For the precise calculation of these picking parameters, KUKA partnered with vision expert Roboception. They develop a technology package that’s easy to integrate into the production process with no major programming effort.

Picking made easy: with a bin-picking cell and the right software, robots have no trouble reaching into various bins.

Sophisticated Path Planning for Collision-free Bin Picking

Specifically, Roboception’s 3D camera is included in the technology package. It sends images of items for picking and their environment to a computer for perfect, singularity-free path planning. Meanwhile, the KUKA.SmartBinPicking software calculates the most efficient collision-free path for the robot arm to reach the desired item. It takes the robot structure and the gripper into consideration while doing this. Moreover, the collision-free path planning reduces downtime to a minimum. When detecting objects – whether sorted or unsorted – the software also prioritizes parts that the robot should grip first, perhaps because they are on top.

How to reach into any bin: automation with 3D machine vision.

AI-based Learning Process Optimizes the CAD Model

Object recognition is trained in advance in a photorealistic simulation environment that uses a CAD model optimized for color reactions, materials and lighting conditions. In particular, AI optimizes recognition of any object for identification even in difficult scenarios. “Each object to be picked from a bin is trained over a specific time period in our AI-supported simulation environment. This training process can run overnight so the application is ready for use the very next day,” said Michael Hohenäcker, Portfolio Manager for Handling and Vision at KUKA.

High-precision bin picking for any objects: a sensor allows the robot to recognize the workpiece quickly.

Intelligent Bin Picking for Even More Automation

Also, the collaboration with Roboception has enabled KUKA to offload path planning to an external computer with no impact on robot controller performance. Hohenäcker explained that this offers a very specific advantage: “Even while the robot is bin picking, the next path is already being planned in parallel, which reduces cycle times.”

Moreover, customers also save time when they integrate the solution. With modern web-based configuration technology, customers can view their unique case in a 3D model. They can control system specifications with ease, with no need for in-depth programming experience.