In Part 1 and Part 2 of our series on Selecting a Part Feeder, we explored the advantages of two different part handling systems: flexible feeders (reconfigurable 2D part handling) and bowl feeders (continuous, high throughput handling configured for a single part).
In the next part of this series, we examine a vision-guided part handling system with even greater versatility: 3D bin picking.
Bin picking enables manufacturers to automate repetitive manual part handling processes to achieve higher throughput, accuracy, and production uptimes, while preventing human exposure to repetitive stress injuries and hazards from handling heavy or high-risk parts (sharps, chemicals, temperature extremes, etc.).
Due to the rapid evolution of part feeding technologies like vision sensors, image processing, robots, and vision guidance systems, 3D bin picking is evolving as an increasingly viable solution for complex part handling useful to both large and small automators.
And, because of the precision achievable by articulated robots, 3D bin picking is a versatile solution for diverse applications ranging from heavy industrial parts to lightweight injection molded plastics, sensitive electronic components, delicate medical parts, and much more.
To find out if it’s the right part handling solution for your manufacturing automation environment, let’s explore the various advantages and challenges of a 3D bin picking system.
What is 3D bin picking?
3D bin picking is a method of pick and place (PnP) part handling that uses a vision-guided, six-axis robot to relocate a single part from a bulk bin of disordered parts. A part is retrieved from the bin and placed onto the next stage of a manufacturing process.
A 3D bin picking system is typically comprised of the following components:
A) Bulk parts bin – Container of parts staged under a vision-guided robot; parts can be stored in the bin at random or in ordered stacks
B) Pick and place six-axis robot with tooling – Programmed to move safely within a predefined space above, in, and around the parts bin; receives coordinate data from the vision system to precisely retrieve a single part from the bin and move it to a predefined location for further downstream processing; appropriate end of arm tooling (EOAT) is determined by where and how a part needs to be picked and placed, such as picking from the inner or outer diameter of a part like a bearing to match its corresponding placement over a shaft or into housing
C) Area scanning vision sensor – High resolution camera installed above the bulk parts bin or on the robot arm; scans the bulk bin of parts each cycle to identify optimal position and orientation of the next single part to pick and place; vision sensors are typically stereoscopic or are combined with displacement sensors to establish depth information about the part arrangement
D) Destination stage – Location where parts are staged for further downstream processing, often a conveyor, nest, or workcell
E) Vision guidance system – Vision software interface connected to sensor/camera and robot configures part recognition patterns and processes images to identify part location and orientation for the PnP robot; software is typically installed on an industrial PC with human machine interface (HMI) display or may be supplied as a self-contained system from the supplier
Precision Robot Integration & Part Retrieval
Random part picking from a bin requires precise articulated movement through three-dimensional space like a human. A six-axis robot is required to achieve a full range of motion in all six degrees of freedom (6DoF) to move, orient, and rotate through the XYZ positions and three rotational axes.
Integrated with a sophisticated vision system, the robot is able to locate a part at precise coordinates regardless of its location, depth, or orientation in the bin. Using a suitable mechanical or pneumatic gripping tool, the robot’s end effector can retrieve and relocate nearly any part type, geometry, size, or material.
Bin picking solutions need to generate approach vectors that include the specific geometry of the bin, robot, and end of arm tooling to ensure no collisions occur when approaching the part. Although the vision-guided robot must be taught to complete all movements and tasks without damaging the part, bin, environment, or humans, robot controllers are becoming increasingly easier to configure. Some developers have designed more intuitive interfaces and quicker walkthroughs to teach the robot its coordinate and reference frames.
Advanced Vision & Part Recognition
An advanced vision system is configured to identify parts based on pattern recognition. Some systems build up pattern recognition by an iterative scanning process of individual parts; others simply use an uploaded 3D CAD model of the part. Once the pattern is defined, a high-resolution vision sensor captures a single image of an entire bin of parts and quickly differentiates the optimal position and orientation of a single part. Image detection and processing is achievable in <0.5 seconds before triggering the vision-guided robot to retrieve the part from the bin at the specified coordinates. With newer bin picking technologies, like Apera AI, these speeds can go down to 0.25 seconds – see below.
Generally speaking, a single part-to-part cycle, from the vision scan of the bin, part detection, robot retrieval, and relocation at the destination stage, can be completed as quickly as 4 to 10 seconds. Cycle speed is dependent on multiple factors, such as the complexity of part geometry, size, and mass coupled with the technical performance of both the 3D vision and robot systems.
The underlying software engineering and algorithms that enable high-speed part identification and relocation are complex, but developers continue to improve their user interfaces for quicker setup and easier use. Thanks to simplified interfaces and advances in machine learning, configuration of a 3D bin picking system is a more approachable and feasible solution than ever before for a broader population of automators.
Explore: Demonstrations at right from OMRON and Pickit 3D.
AC helps you select and integrate the right 3D bin picking solution for your unique application requirements.
RESOURCE CREDIT: OMRON | Bin Bicking Solution for Flexible Manufacturing
RESOURCE CREDIT: Pickit | Fast bin picking with Pickit 3D and ABB
What are the tradeoffs of a 3D bin picking system?
3D bin picking offers many strategic part handling advantages, as well as a level of complexity worth considering before investing in a solution.
Advantages of 3D bin picking
> Ideal for repetitive PnP of a single part from a bulk container
> Accommodates fast, simple, software-based component changeovers; supports many different component models
> Stable, predictable, and efficient part handling for nearly any part size, geometry, surface characteristics, and mass
> Highly configurable system and tooling
> Highly adaptable for different applications and types of parts
> Safe to operate; prevents risks and repetitive stress injuries to humans
> High uptime once configured
> Footprint is relatively small for the robot, but a larger space is required for manufacturing to accommodate part staging, bins, robot, vision platform hardware, conveyor or other staging requirements, etc.)
> Cost-effective for repetitive part handling; low total cost of ownership (TCO) over lifetime performance due to stable and predictable performance, high uptime, and low maintenance
> Next generation AI-driven bin picking technology offers affordable and adaptable solutions for picking more than one part (see below)
Challenges of 3D bin picking
> Higher initial capital investment than 2D systems due to advanced technology requirements
> Requires human configuration of parts using the vision system, though user interfaces are increasingly designed for simplicity and adjustability from 3D CAD part models
> Requires the robot to be configured for its environment, integration with vision system, and fine-tuning for optimal speed and throughput
> Fully autonomous operation requires more planning for system integration to ensure the efficient and error-free handling of parts throughout the manufacturing stages and processes
What is the right part type for 3D bin picking?
Because of high-resolution vision imaging and highly configurable end effector tooling capable of different gripping or pneumatic suction mechanisms, 3D bin picking is highly adaptable to diverse part geometries, sizes, materials, and surface characteristics.
The biggest challenges are centered on vision part identification and robot adaptation for asymmetrical or entangled parts in the picking stage. Though advances in vision machine learning and AI enable quicker resolutions for these conditions, it’s important to consider what parts work best with 3D bin picking.
Parts that work well with 3D bin picking
> Parts of diverse sizes and masses, from lightweight and delicate parts like syringes and vials to large and heavy automotive parts
> Parts made from a diverse range of materials, including metal, plastic, rubber, and textured parts (like electronics boards) or parts with soft or slippery surfaces (like cylinders made of glossy or shiny plastic)
> Parts that are not likely to be damaged or entangled when stacked in a bulk bin
Parts to avoid with 3D bin picking
> Parts that can easily entangle with each other or clump together (e.g. asymmetrical shapes, magnetic parts, etc.)
> Parts that are difficult to differentiate by vision scans under adequate lighting conditions
> Parts that are too small or too large to be gripped mechanically or pneumatically by the robot end effector tool
> Parts with a top and bottom orientation that require a flip operation when presented bottom-up in the bin
Are there ways to enhance a 3D bin picking system?
A typical 3D bin picking system includes all the components required to enable its powerful vision-guided pick and place functionality. However, depending on the application, there are additional components or upgrades that can add value and boost performance, including the following:
> Integrated robot controllers for managing high-speed machine motion control
> Part staging enhancements, like a robot slider for switching from empty to full bins
> Custom end effector tooling to enable PnP of moving parts
> Vibration unit that shakes bins to enable a faster rate of part removal
> Bin agitator to re-orient unpickable components
> 3D sensor upgrades for higher resolution of point cloud data or for IP65 protection from environmental risks (temperature extremes, particulates, residues on parts)
> Robot workcell features required for safety compliance requirements (barriers, guarding, light sensors, and safety curtains)
What innovations are improving bin picking technology?
We asked AC’s Senior Application Engineer for Automation, Jerry Entrikin, to describe the latest bin picking improvements. He pointed to Vancouver-based, Apera, and their AI based “4D Vision Technology” technology, defined as “2D Vision + Ai.”
Their proprietary Apera Vue software uses an AI neural network to determine the best movement path and to direct the robot to pick and place objects quickly and precisely achieving the fastest cycle times in the industry, down to 0.25 seconds. The blending of machine vision, AI, and neural network part recognition is a true leap forward.
Apera’s Vue vision software can also analyze a 3D CAD model using machine learning to create a virtual simulation of the proposed robot PnP workcell. This information makes it easy for integrators like AC to predict performance more accurately which helps us de-risk projects for our clients.
Jerry shared just a few of the strengths he noticed about Apera’s AI based solution:
“The industry is evolving quickly and bin picking solutions from Apera AI appear to the be the wave of the future. Their AI based solution quickly identifies parts that are reflective or translucent under highly variable lighting conditions – all of which are problematic for conventional machine vision solutions.
When bin picking more than one part, Apera AI enables additional camera pairs that can be added to the same vision controller at a lower price that conventional feeding methods that would require two feeders. Their solution is now on par with the cost of vibratory feeding for the same parts.
It’s also ideal for high product mix bin picking that requires rapid changeovers selectable within the software; hardware reconfiguration is no longer required to changeout parts for the picking operation.
All together, with the increased accuracy and speed of part detection, it’s a significant improvement for bin picking.“
Explore: Apera’s 4D Vision delivers significantly improved part detection and cycle speeds.
RESOURCE CREDIT: APERA.AI YouTube Channel
How does a test plan impact implementation?
Because 3D bin picking relies on complex vision hardware, software, and an integrated six-axis robotic system, a thorough test plan is crucial to ensure parts are configured for quick and accurate identification and retrieval.
Implementation Objective
Testing Considerations
Part Recognition: Ensure the vision system accurately detects parts.
Does the part need to be scanned or uploaded into the vision system?
Does the part pattern need to be adjusted in the vision system for better registration?
What adjustments are needed to increase vision accuracy for parts with higher reflectivity, transparency, or textural challenges?
Vision System: Ensure high-end vision detection of parts
Is the vision sensor/camera at the optimal physical location to capture the area scan of the parts bin?
Is the position of the vision sensor/camera adjustable, such as on the robot arm, or at a fixed position above the bin?
Will the light projection for the vision system enable accurate part identification? Is additional lighting required?
Vision Sensor Location & Lighting: Ensure the camera position and lighting conditions are optimal for successful detection of parts
Is the vision system appropriate for the required resolution, speed, and accuracy regardless of part geometry, shape, and position?
Are multiple vision technologies supported for 3D mapping, 3D laser projection, 2D detection, and code scanning?
How many cameras and resolutions are supported by the vision system?
Are automated calibration tools included to simplify manual calibration?
Does the vision system support part identification in non-moving and moving stages?
Does the vision system capture the full range of motion required for the robot?
Are vision components upgradeable? Is the system scalable?
Does the vision system use deep learning or AI advantages that improve the speed and accuracy of part identification?
Robot Performance: Ensure the robot is appropriate for the application and configured correctly for the environment, bin location, and part position and orientation
Does the robot support the type of part retrieval required for the application?
What adjustments are needed to adapt to complex part positioning, clumping, or entangling?
Can the robot integrate with the vision system? Does the vision system capture the full range of motion required for the robot?
Does the robot deliver the movement speed and cycle times required for the application?
Are there automated tools that make manual configuration of robot frames quicker and easier?
Does the robot perform efficiently and with high repeatability? What adjustments are need to optimize performance?
Can the robot be quickly repurposed for different parts in different locations in the environment?
Do the application specifications require robot safety barriers, guarding, light sensors, or safety curtains?
AC expertise to make the right decision
> Are you ready to automate a time-consuming and labor-intensive part handling process?
> Does your part handling throughput need better performance and repeatability?
> Do you need engineering expertise to support your existing 3D bin picking system?
AC has extensive automation experience with multi-axis robot and vision system technologies, as well as complex part handling systems. We can help you land on the right part handling solution and determine if automation makes sense for your application.
How can we help? Talk with an engineer today.
