Fully automatic picking of unknown products from bulk material

Application

The vision system acts as the eye of the robot. It detects grippable surfaces on the objects to be processed in the source container and calculates grasping points and collision-free paths for the robot. This enables the robot to pick up unknown products directly from bulk material and place them in the target bin in the area with the lowest filling. Vacuum cups ensure gentle gripping of the respective objects. A previous "teach-in" of the individual products is not necessary. Regardless of whether it's a medicine bottle or a tea pack, the multifunctional gripper can be individually designed for a wide variety of articles with the most diverse packaging units. The system learns over time which of the different graps works best for each item. The achievable picking performance for a process-safe system is highly dependent on the characteristics of the gripping objects and ranges between 300 and 500 parts per hour.

Two Ensenso N35 3D cameras provide the system with the necessary image data. All Ensenso 3D cameras work according to the "Projected Texture Stereo Vision" method. Each model uses two CMOS sensors and a projector that projects high-contrast structures onto the object to be captured - even in difficult lighting conditions. Ensenso cameras operate using stereo vision, which imitates the human vision. The result is a 3D point cloud as the basis for the required spatial object information.

To integrate the cameras into autopick, psb intralogistics used the Ensenso SDK. In addition to wizards for easy setup and to support camera calibration of the 3D cameras, it includes the option for GPU-based image processing for even faster 3D data processing.

It also enables the output of a single 3D point cloud of all cameras used in multi-camera operation, which is required in this case, as well as the live composition of the 3D point clouds from multiple viewing directions.

The first camera is installed above the source box in order to consider the latter as a collision object in the path planning of the robot arm. Here the unknown parts are presented and the point cloud for searching the appropriate handle point is generated. The latter is done with the help of the Mikado ARC (Adaptive Robot Control) software from isys vision. It combines Ensenso's 3D stereo vision camera technology with an easily configurable, adaptive robot controller. The result is a complete 3D robot vision solution for bin picking and parts handling with an autonomously operating robot, as supplied by autopick. Instead of following predefined, taught and firmly defined paths, it orients itself independently in the workspace and reacts to every situation.

The second camera is above the target box. It searches for free storage positions. In addition, the Z-height of the box content is determined in order to be able to take this into account when determining the storage position. It is crucial that the target box is occupied evenly and that the products are placed carefully. The latter is particularly important when picking fragile parts. The image acquisition is asynchronous to the movement of the robot in order to optimize the cycle time.