!!! Attention Store Owners and Collectors !!!

For over 70 years, tens of millions of children and adults have enjoyed creating and building with those very recognizable and wonderful interlocking pieces of molded plastic. But having hundreds, or thousands, or millions of those pieces in a collection or inventory poses one problem: how to find the piece or pieces you're looking for, and how best to organize your collection for searching?

There are essentially two ways to deal with organizing your collection: tediously search through it piece by piece, or categorize everything to streamline the process -- Sorting. This task ultimately falls to you, friends and family, or paid helpers. Yet, finding the time, patience, and budget for necessary sorting is too challenging for most. Manual sorting often prompts a sigh of despair, a stark contrast to the happiness the collection or inventory provides. Manual sorting is undertaken out of necessity, not desire, due to the lack of feasible alternatives.

In addition, manual sorting can account for up to 70% of overheard costs. Imagine virtually eliminating that expense! Imagine further the idea of actually keeping up with your bulk sorting!!! For collectors, more sorted inventory means more time to build! For store owners, more sorted inventory means more sales!

This ridiculous problem has demanded a solution for over 70 years! Finally -- the solution is Fast and Accurate Automatic sorting!

Third Arm Robotics is happy to announce the launch of the Standard Sorter v1.0. This working prototype industrial robotic arm and software package will automatically sort totes of your favorite little pieces of molded plastic into all of the different colors, sizes, and types specified by you! Object type identification (an AI model) opens up the opportunity for automatic sorting by part number, search for specific pieces, parting out, automatic cataloging, sorting by set membership, custom sorting, etc. Even though it has an accuracy of greater than 90%, this system is not yet a total replacement for manual sorting. TAR continues to improve this accuracy, but until then the Sorter will be a very useful and valuable assistant in your sorting process.

In the next few months there will begin an opportunity for individuals and companies who are serious about sorting to be part of the solution.

Ready to see the Sorter in action? Check out these Demonstration videos:

1. Standard Sorter v1.0 General Demo

2. Standard Sorter v1.0 Color Demo

3. Robots are Persistent (2023.01.16)

4. Sorting Demonstration (2023.01.23)

5. Different Style of Object (2023.02.09)

6. Sort by Object Type Demo (NEW 2024.04.04)

(More videos will be posted here as they're produced)

What am I watching in these demonstration videos?

You are watching the Standard Sorter Version 1.0. On the left is a view of the robotic arm, sorting bins, dispenser, and part of the working area. On the right is a close-up view of the entire working area. Mounted above the working area and not visible in these videos is a digital camera.

How is the sorter doing this?

That is a very complicated answer. In short, the system uses a robotic arm, computer vision, and machine learning to sort all the objects it sees. The idea is that the system will 'look' into a tote/box/bin of objects, find one of those objects, and then using the arm: grab it, move it, and place it into a predesignated sorting pile/container.

What kind of Robotic Arm is that?

It's a Dorna 2 industrial robot.

How many moving parts are there? Is this system very complex?

There are two moving parts: the robotic arm and the dispenser. I've intentionally kept the hardware very simple so it's easy to build and easy to use, with little chance of failure because there's not too much that can. It also has a pretty small footprint (about 1 square meter) so it can fit in almost anywhere.

The sorter is already pretty fast, but can it go faster? Can it sort as fast as a human?

Yes it can sort faster than what is demonstrated. I'm still learning how to properly operate this robot, so the speed you're seeing is a 'safe' speed for the hardware. Not including overhead for object management (also demonstrated in the videos), each object currently takes the sorter approximately 7 seconds to find, grab, move, and place into a bin. Humans are very good at sorting and will probably be faster than robots for a little while longer. The Sorter makes up for this speed difference by being able to operate 24/7, which humans can't (or probably shouldn't) do.

Why does the sorter miss grabbing some objects, and why are you showing that in your demonstrations?

The quality of the grabbing depends on how well the system can see the objects in the working area. For various reasons, sometimes objects just aren't seen very well and the object gets missed. The demonstrations are showing typical operations, not extraordinary operations (which IMO would be dishonest). I will not over-sell or under-sell this system. It's very important to me to be completely honest about the capabilities of the sorter, making sure that everyone understands what it can and (most important) cannot do. My primary motivation is to solve a frustrating problem that's been around too long.

How well does this system grab objects?

The system will grab every object it sees in the working area. Single-try grabbing accuracy is currently 85%. Please note that even though it misses an object once, it'll very likely grab it the next time it tries.

Why doesn't the sorter find and grab all the white objects at the end of the general demonstration video?

With the white background in the working area, it's sometimes difficult for the system to detect all of the white objects. Future versions of the system will solve this problem.

Will this do all my sorting?

The Standard Sorter is not a total replacement for manual sorting -- yet. For now, please think of it as assisting in the sorting process -- freeing up some of your time and other resources.

Can this system sort by object type?

YES!!!! The current version of the Sorter will sort (with an accuracy of about 90%) a limited number of object types, including bricks, plates, slopes, technic bricks, additional technic components (like technic gears and connectors), missed grabs, and unknowns. In future upgrades the model will be able to identify 15 different object types and 100 part numbers.

How well does this system separate colors?

The system currently separates colors with an accuracy of 94%.

Don't we all agree that we shouldn't be sorting by color?

It depends on the situation: For small collections/inventories, sorting different types of objects by color is manageable. Larger collections/inventories usually start sorting by object type, and then sort each object type by color. In either case, sorting by color is still needed. Remember that the sorting service I'm offering should be used as an assistant in the sorting process, but not (yet) a total sorting solution.

How can I get one of these so I can offer a sorting service, or sort my own collection/inventory?

Soon there will be an opportunity for individuals and companies to acquire and operate their own sorter. The opportunity page will keep you updated, and you can always email me if you have any additional questions.

How expensive is this? Can I justify the capital investment?

Industrial-grade hardware for these systems isn't cheap (literally and figuratively), but the price definitely isn't out of reach by anyone serious about sorting. The periodic license fee will be negligible compared to manual labor costs. To get a decent ROI, this system is for anyone who currently spends more than 20 hours a month on manual sorting. Run the numbers: Should I Consider Automated Sorting?