Robotic bin picking defines different processes where robots are utilised to pick and place materials from a bin. Bin picking is a tricky robotic task in the robotic realm. However, you don’t have to solve it using an intricate solution. Bin picking using robot vision can be executed in both an easy and a hard way.
In the last few decades, robotic experts have conducted massive robotic research and development. This has resulted in enhanced robots capable of detecting disorganized items in a box and plucking them individually.
While this task is easy for humans, it has presented challenges when it comes to robotic manipulation; including artificial intelligence, multi-fingered grasping, trajectory planning, and robot assembly vision.
Today, bin picking is more possible than it was a few years ago. Robot advancements over the years have solved many of the limitations which hindered bin-picking applications. Still, there’s a lot to be done.
Even with the advancements, many existing solutions are intricate and need robust technology such as advanced three dimension configurations and machine learning. However, Universal Robot experts opine that robot users don’t have to adopt the complex solutions to achieve success in bin picking.
Bin picking presents various challenges for robot vision. They include:
The items create shadows upon each other and this camouflages them from the camera.
Some items are either completely or partially camouflaged by the items on top.
Detecting the beginning of one object and the end of another is difficult. This presents difficulties when it comes to recognizing the layout of each item.
These challenges are available both in three dimensional and two dimension vision. However, they are more challenging in two dimension robot vision and can present extreme difficulties when it comes to recognizing individual items.
Executing Bin Picking the Hard Way
There are different commercial bin picking solutions which are intricate, expensive, and inflexible. You will need to include additional components to the robot for the system to function effectively. They include lighting and fixed cameras.
Many of the available solutions utilise three dimension vision which requires more processing. Further, users need exclusive technology to achieve success in these solutions.
An exemplary bin picking configuration can include:
- 3D laser scanner
Three dimension laser scanners utilise light to record three dimension depth images. The sensor generates a point cloud(a collection of data points) of the items and the surrounding environment.
- 3D object detection
3D object detection strives to discover items within the three-dimensional point cloud. Some of them utilise CAD versions of the items being recognized.
- Stereoscopic vision sensors
Stereoscopic vision sensors utilise dual cameras to generate a three-dimensional image of the surroundings. You can utilise them alone or incorporate a laser scanner to enhance detection precision.
- Fixed lighting
Some systems need additional lighting to consistently brighten the scene.
The above systems vary in terms of complexity. While some are tiny and independent units firmly fixed above the recognition area, you will need to configure a variety of lights and sensors in specific locations within the surroundings in other systems. Regardless of your preferred method, you will need more technology other than a robot assembly camera and this can be costly.
When the technology has been configured and the system trained, the recognition can be powerful. You will need to program the robot in order to enable the sensor to relay item locations and proceed to utilise the trajectory planning to pick up the items.
Executing Bin Picking with Ease
In this tactic, you only need a simple robot camera to accomplish intricate bin picking with less intricate sensing technology. Utilise the robot gripper to pick a handful of your preferred items. In this technique, item detection is not necessary.
All you need is to place the gripper in the box and pick. Place the items on a flat surface and let the robot vision sensor recognize individual items on the surface. Now proceed to select each item one by one to complete the process.
While many robot users may prefer the easy method, it is not ideal for every application. Users should have items that don’t require picking in a similar manner all the time. If your items are fragile, they will require delicate handling which means that you will not pile them in a bin at all.