Reduce the burden of deploying a 3D robotic bin-picking system.
Manufacturers and systems integrators face numerous challenges when it comes to deploying robotic bin-picking systems. Robotically picking objects from cluttered bins or shelves for packaging, assembly, machine tending, and other complex manipulation tasks has proven difficult for nonexperts to deploy. However, advancements in 3D imaging technology and artificial intelligence (AI) software have now made it possible for technicians to implement bin-picking systems on their own factory floors.
When a company properly designs and implements a 3D-vision-guided robot, it can significantly enhance its operations. Moreover, the burden of deploying a bin-picking system has been greatly reduced thanks to advancements in underlying technologies, from robot arms and 3D imaging systems to AI software that combines optimal math-based motion planning and geometry algorithms with intuition-based machine learning algorithms that are more humanlike.
3D machine vision
To differentiate and pick loose, randomly oriented objects, a vision-guided robot must complete various complex steps. These include capturing the objects with a 3D camera, analyzing the parts in software, and understanding their orientation. The robot needs to receive information about an object’s position, gripping points, and optimal movement path to execute the pick. When objects are piled randomly, the software must determine which object to pick, even when parts are only partially visible.
Additionally, when parts or objects are tangled, the robot arm might have to perform complex rotations or put the part down and regrasp it in the correct orientation. CapSen Robotics solves these problems with proprietary AI software that combines classic CAD matching techniques with proprietary machine learning, 3D vision, and motion planning algorithms. The software utilizes 3D models of objects that can either be provided by the customer or scanned-in using CapSen Scanner.
CapSen’s PiC AI software employs geometry-based vision algorithms and machine learning techniques to continuously improve object detection accuracy. By utilizing 3D models, the AI software can discern how to accomplish the manipulation task at hand, which often requires holding and positioning a part in very specific ways so the part can move onto the next step of the process—assembly, packaging, or something else.
30 parts per minute
CapSen’s PiC software offers consistent and complete control of the robot, end effector, and 3D cameras within a robotic cell, significantly reducing the integration burden of deploying a bin-picking system. Additionally, its image processing and planning times of less than a second make it the fastest software available, enabling PiC-powered cells to pick-and-place up to 30 parts per minute.
The hardware-agnostic AI software tackles the most challenging bin-picking problems by combining 3D vision with advanced motion planning and control software, providing robots with spatial intelligence without the need for custom programming for new parts. It also ensures collision avoidance and enables advanced motion planning tasks, such as detangling and assembly.