
Menu
Moonflower
what is moonflower
Bin picking, what in Italian is called “presa da cassone” or in German “Griff in die Kiste” is a process for loading machines thanks to robots and 3D vision.
Moonflower is our random bin picking software solution. It is used for product localization inside a bin and it generates a picking trajectory for reaching the product. It integrates with the robot program in order to efficiently solve complex localization and picking problems with high flexibility.
Products are localized using 3D data provided by one of the supported scanners: fast and robust matching algorithms are exploited to find the 3D position and orientation of an object starting from its CAD model.
Moonflower generates a picking trajectory starting from the pose of a localized product, checking for collisions between the robot and the surrounding environment.
In addition, when the shape and distribution of the products inside a bin are very complex, an automatic extraction strategy can be defined for a product model to allow the robot to exit the bin with the grasped part in a cleaner way.
what can i do with moonflower
Moonflower is commonly used to drive a robot that serves two turning centers at one time picking from two bins with different parts. Depending on layout and cycle time the system could use two cameras or a single one moved from one box to the other. A flexible gripper makes model switching almost instantaneous. Everything is programmed easily from a 3D cad file.
A special part model is designed to easily pick metal sheets, even when the thickness is a fraction of millimeters. Starting from a 3D or 2D drawing (.dxf, .geo, .hpgl) and quickly defining how to grip and how to place a file robot in a CAD file, it is possible to load instantly any machine. A plugin for palletizing a bent part is also available.
Sometimes a robot is already installed to pick parts from a conveyor at a fast pace, so all you need is to fill the conveyor with parts from the bin.
In this case, you can use Moonflower with a cost-effective 3D camera (usually ToF) and a gripper that can take a bulk of parts (often a magnet) to empty efficiently the bin.
Some Moonflower advantages may be critical for your application with a vibrational unit: a very effective way to distinguish flipped parts, accurate picking at different heights, and definition of multiple picking positions from a CAD file.
Billet’s model can be automatically generated from parameters for both cylindrical and squared section types. Often cycle time allows a camera on the robot end effector so several bins could be served with no additional costs.
Moonflower is commonly used to drive a robot that serves two turning centers at one time picking from two bins with different parts. Depending on layout and cycle time the system could use two cameras or a single one moved from one box to the other. A flexible gripper makes model switching almost instantaneous. Everything is programmed easily from a 3D cad file.
A special part model is designed to easily pick metal sheets, even when the thickness is a fraction of millimeters. Starting from a 3D or 2D drawing (.dxf, .geo, .hpgl) and quickly defining how to grip and how to place a file robot in a CAD file, it is possible to load instantly any machine. A plugin for palletizing a bent part is also available.
Sometimes a robot is already installed to pick parts from a conveyor at a fast pace, so all you need is to fill the conveyor with parts from the bin.
In this case, you can use Moonflower with a cost-effective 3D camera (usually ToF) and a gripper that can take a bulk of parts (often a magnet) to empty efficiently the bin.
Some Moonflower advantages may be critical for your application with a vibrational unit: a very effective way to distinguish flipped parts, accurate picking at different heights, and definition of multiple picking positions from a CAD file.
Billet’s model can be automatically generated from parameters for both cylindrical and squared section types. Often cycle time allows a camera on the robot end effector so several bins could be served with no additional costs.
what is bin picking?
Strictly speaking, the bin-picking problem involves the picking of a part from a box where many items are mixed.
In an industrial application anyway, we have to extend normally the analysis to the subsequent placing of the object somewhere else.
Placing can be really critical when analyzing the project since it may add constraints in gripper design or create a deadly difference in cycle time.
It is common to refer to “Robot Vision” by Berthold and Horn (MIT, 1986) as the beginning of any discussion about bin picking in industrial applications.
Since then, often announced as an incumbent revolution, robot bin picking for a long time has been limited by the lack of 3D sensors with the necessary resolution and speed, computational power, and experience in solving the pose estimation and path planning problem.
It is evident that while software is, at least in principle, universal and a sensor can address a really wide range of parts robot end effector has usually to be designed each time. Robot grasping with contact compliance is still not evolved enough to allow the use of hand-like end effectors at real application speed.
For sure this need for customization can reduce the flexibility advantage of a bin picking system compared to other loaders but if this is the main reason for slow market growth we should see a wider range of applications in cases where a gripper device is easier to design.
For example, many classes of machines are loaded with quite symmetrical objects (such as billets in an oven for hot stamping or parts for turning centers) and it is possible to create a general gripping system for them. A second clear example is a system that has to load a limited number of objects (sometimes only one) for its complete lifetime.
All semistructured bin picking have also easier gripping requirements since only a very limited number of poses are going to be handled. So there is for sure a market much larger than the one currently addressed even giving a maximum weight to the end effector design challenge.
A bin-picking system is built by providing a robot workcell with:
A sensor to build a 3D image
a software that finds parts in a 3D image, calculates a safe pick position if it exists and, plans a path there avoiding collision
a gripper that can reach parts in a sufficient number of poses to ensure the box can be emptied
a robot arm to perform the path.
MOONFLOWER SENSORS
The general output of a 3D camera is a point cloud. Current 3D sensors are usually divided into four groups: stereo cameras, structured lights, laser triangulation, and time of flight.
Euclid Labs is sharing data about the best sensors available www.industrialrobotics.org.
Moonflower supports a lot of different 3D vision cameras to provide the best solution for performances and prices at the many automated bin picking tasks robot integrators and OEM are facing today.
Some examples of applications are:
moonflower Gripper
Vacuum systems, magnets, and mechanical grippers are all suitable for bin-picking applications and in some cases, they could be used simultaneously.
A bin-picking gripper has to face some unique challenges:
Failing the gripper design will lead to deadlock conditions where parts are localized but there are no available picking positions.
If these conditions happen only on the low corners it may be an accessible limitation of emptiness efficiency.
Moonflower supports multiple grippers and multiple tcp in each gripper.
moonflower Robot
A lot of different robot brands are available and many of them are distributing arms with different kinematics. In Euclid Labs bin picking software 6 axis, puma-like and not, robots are available, external axes are optional (a linear unit may be a very practical way to extend bin numbers, usually coupled with a camera on the robot), 4-axis palletizers are supported for easy situations where pose is fixed.
It is often undervaluated how much selecting a higher payload robot has an effect on cycle time. In this video, cycle time moved from 6.5 sec. to 7, 7.8, and 12.2 sec.