AI Venture from the University of Tokyo Started providing “ANY PICKER”, a picking system that enables high-precision and high-speed bulk picking without a model

[AI Venture from the University of Tokyo] Started providing “ANY PICKER”, a picking system that enables high-precision and high-speed bulk picking without a model
– Contributing to labor saving for picking tasks at every site –
TRUST SMITH Co., Ltd. (Bunkyo-ku, Tokyo), which provides solutions utilizing cutting-edge AI and robotics technology, provides a picking system “ANY PICKER” that enables high-precision and high-speed picking of bulk without model. started.
This product enables high-speed (recognition in 0.1 seconds) and high-accuracy (99.99%) picking without prior learning, and greatly contributes to labor-saving for picking tasks at all sites.
[Image 1

Picking Issues at Manufacturing and Logistics Sites
The shortage of labor due to population decline is a serious issue in manufacturing and logistics sites, and automation and efficiency improvements using robots and AI are required.
Among them, the replacement of picking work by robot arms is the cornerstone of DX in factories and warehouses.
Conventional picking software requires pre-registration or
pre-learning of picking objects, and cannot pick objects for which data cannot be prepared in advance.
For example, it was difficult to pick irregular-shaped items such as fried chicken at EC distribution warehouses where there are tens of thousands of products to be handled, and food factories.
TRUST SMITH has developed a unique picking algorithm that solves these issues. Features of modelless picking system
TRUST SMITH’s modelless picking system has the following features and greatly contributes to labor saving at manufacturing and logistics sites.
1. Picking is possible without teaching
2. Supports difficult-to-recognize objects such as amorphous objects 3. High speed picking
4. Highly accurate picking
Feature 1. Picking is possible without teaching
Picking is possible without prior learning, so it is possible to pick objects that are difficult to learn in advance.
In addition to being able to adapt to the site where the products handled change frequently, it does not require a learning period, so it can be introduced in a short period of time without technical understanding.
Feature 2. Compatible with objects that are difficult to recognize, such as irregular objects
[Image 2

Bulk objects to be recognized (including amorphous objects such as detergent refill containers)
[Image 3

Succeeded in picking by calculating the adsorption point even for irregular shaped objects.
It is possible to deal with the following objects that were considered difficult to recognize in the past.
・Unshaped items (deep-fried chicken, detergent refills, etc.) ・Objects with metallic luster or plastic luster (metal parts, aluminum foil, etc.)
・An object such as a box that is laid out on a flat surface without gaps (*1) ・Rounded objects (cup noodles, etc.)
[Image 4

Succeeded in picking by calculating the suction point even for rounded objects. As shown in the image below, it is difficult to recognize the boundaries of tightly packed boxes using only depth images, but this system detects object boundaries from RGB images and excludes candidates that are difficult to attract.
[Image 5

Recognition image of boxes covered
Feature 3. High-speed picking
The pickup position can be calculated in just 0.1 seconds, realizing high-speed picking.
As a result, it can be put to practical use even in the field where the takt time is strict.
Feature 4. High-precision picking
As a result of the picking test (*2) targeting 50 items, a 100% adsorption success rate was recorded.
*2: 50 items with various shapes and properties, such as cup noodles, detergent refills, cosmetics, spray cans, iPad product cardboard, etc., were tested under the same conditions (containers, lighting, etc.).
technical details
The flow from recognition to adsorption in this system is roughly as follows. 1. Acquire depth image and RGB image of adsorption target with depth camera and RGB camera
2. Fusion process (*3) the depth image and RGB image to calculate the final position
3. Route planning/adsorption
*3: Fusion processing refers to combining and processing multiple different types of image data.
Application range of modelless picking system
[Image 6

The scope of application where this system can contribute to labor saving is as follows.
The task of packing the items transported from the automatic warehouse of the EC distribution warehouse, where the products handled change daily, into the same cardboard for each order.
1. Stuffing tasks such as lunch boxes in food factories and
supermarkets that handle irregular items such as fried chicken 2. Picking bulk screws with metallic luster and multiple sizes and shapes 3. Picking workpieces of different shapes, such as crops and marine products 4. Picking in unmanned environments such as outer space
Future prospects
[Image 7

TRUST SMITH Co., Ltd. will continue to further improve the modelless picking system “SMITH PICK”, as well as the next-generation automated warehouse system “RENATUS”, AGV optimum control system “PYUTHIA”, general-purpose transport robot “Kaghelo”, etc. By combining with the products owned by the TRUST SMITH Group, we will contribute to making our clients smarter.
What is TRUST SMITH Co., Ltd.?
[Image 8

TRUST SMITH Co., Ltd. is a venture company from the University of Tokyo that creates innovation using cutting-edge technologies in the fields of AI, mathematical algorithms, and robotics.
In addition to working on the research and development of our own products in various technical fields, regardless of hardware or software, we provide custom-made solutions that meet the challenges of companies and support research.
Company name: TRUST SMITH Co., Ltd.
Location: 7F Kikuhana Building, 4-1-1 Hongo, Bunkyo-ku, Tokyo Representative: Takuma Osawa
Business: Development of mathematical and physical algorithms (AI) Date of establishment: January 18, 2019
Company HP:
-Click here for press releases on each technology of the TRUST SMITH Group- ・An AI venture from the University of Tokyo has successfully developed an algorithm for an obstacle avoidance arm using artificial
・[AI venture company from the University of Tokyo] Succeeded in developing an automatic transport robot using artificial intelligence to realize a society where humans and robots work together
・[University of Tokyo AI venture] Practical application of an algorithm that detects the optimal position for a robot to grasp an object
・[AI Venture from the University of Tokyo] Started providing consulting services for smart factories
・[University of Tokyo AI venture TRUST SMITH establishes research institute] Established independent research institute BSAI and started offering “PYUTHIA”, a unique technology that realizes AGV group control of several to 1000 units.
■ Inquiries regarding this press release
Name: Takuma Domoto
Email address:
Phone number: 03-6822-3276
Company HP:

Details about this release:

Leave a Reply

Your email address will not be published. Required fields are marked *

%d bloggers like this: