Automating Your Processes Through Bin Picking – Here’s What to Look For

Robot Tech

Bin picking

Automating Your Processes Through Bin Picking – Here’s What to Look For

#HowToRobot -
Editorial team
Bin picking has been one of the most challenging processes in automation for years. While the technology can optimize your production in several ways, some tasks may still be too complex for it to handle. This article shows the processes that could be automated through bin picking, the potential challenges, and their solutions.

What is bin picking?

The goal of bin-picking is to pick up similar objects which are randomly placed and oriented from a bin, using a  vision system, a robot, and a gripper. The vision system typically performs a 3D scanning of the object giving  its point cloud (set of points in space that constitute an object) and thus recognizing it.

Then the robot is driven into the bin, picks up the object, goes out of the bin and to the placement target, typically a box or some other fixture.

Challenges and solutions for the robotics part of the bin picking process

Some items will be difficult to reach

One major problem with bin picking is the inability of the robot to pick up all the pieces from the bin. Picking up 80 to 85 percent of them is manageable as the vision system will command the robot to pick up the objects which are easy to grab (e.g the ones sitting on top).

The difficult part is managing to grab the last items which are intertwined or sit in a corner of the bin. This can be solved in two ways.  The first one implies redesigning  the bin, giving it a more practical shape.  For  example, a conical shape would keep all parts in the middle at all times.

A second fix would be to shake the bin (e.g automatically using a motor) so that the parts’ positions and orientations are rearranged and not mixed anymore. The camera would then identify them more easily.

Grippers can damage the collected items

Another problem is the chance of collision of the gripper with the rest of the objects while trying to pick one. The picking strategy needs to be precise in order to avoid damaging the remaining parts (especially fragile  ones).

A solution to that is an accurate path planning algorithm with obstacle avoidance. Path planning in robotics is the process of finding the optimal continuous path from point A to point B. If obstacle avoidance is also included, the same task is performed without hitting any obstacles.

There might be cases where the objects in the bin are of different shape or nature (e.g apples and oranges). In this case, a soft gripper or a vacuum gripper with suction cups could be used so it adapts to any size and shape.

 

Suction grippers
Suction grippers can more easily access corners of the bin or pick up items that are badly oriented. However, this creates a need for rearrangement afterwards.

 

Challenges and solutions for the vision part of the bin picking process

The vision system of a bin picking solution can either include 2D or 3D cameras. It goes without saying that 3D cameras (dual cameras which create a 3D image) are more expensive, but they are more effective. In addition, 3D laser scanners which produce the aforementioned point clouds can be used.

Lightning, occlusion, and edge detection can complicate the process

Recurring problems in vision include lighting, occlusion and edge detection. They apply to both 2D and 3D vision systems.

Lighting means that shadows are cast from each object to the rest so the camera has a hard time detecting them. It can be solved by providing the scene with additional lighting which could be attached next to the camera or on the robot’s wrist.

Occlusion takes place when the object is not fully visible to the camera as another object is  placed on top of it. Again, shaking the bin can be useful in this case, so the object can more easily be detected.be detected easier.

Edge detection is the challenge of figuring out the outline perimeter (edges) of an object. It gets even harder when a high number of objects are placed next to each other like in bin picking.  The fix to this problem depends on the software and the methods used.

 

Grippers can damage the collected items
While picking up an item, grippers are in risk of damaging other items in the near.
 
How to spot potential for automating your processes through bin picking

If you consider automating processes in your company using bin picking, the following points could suggest some potential for automation:

 

  • High number of small parts: If the system is set up properly, the robot can pick up those parts (such as screws or washers) fast and robustly.
     
  • Unordered items: Processes where it is not possible to receive items orderly from a previous process or subsupplier. In most cases, it will be more effective to request items to be delivered in a structured way, and thereby completely avoid bin picking.
     
  • Parts that have a matte surface. If parts are highly reflective, such as machined metal parts, there can be challenges with lightning. Most camera systems will have a hard time finding the items for bin picking.
     
  • A process where the cycle time is more than approximately 10 seconds, and preferably with a buffer for picked parts. It will take a bin picking system some time to find and pick each item.
     

Conclusion

For the developers, bin picking is truly the holy grail. For the customers it may not be so. Usually, it can be a good idea to try avoid using bin picking solutions, simply by ensuring that the delivery format of items is in an orderly fashion, instead of random in a bin.

This will usually add a bit of cost, but it is often cheaper and allow for faster processing than using bin picking.

However, in some cases this is not possible, or very expensive. In those cases, it is great to be able to use bin picking.

If a process of yours ticks all four previous points, it might be very well suited for bin picking. When finding the solution to use, beware that some solutions are still very experimental, while others have had several years to mature and prove their robustness.