Periodic Reporting for period 1 - GROW (General-purpose Robot for Object-retrieval in Warehouses)
Periodo di rendicontazione: 2021-09-01 al 2023-02-28
The current boom in e-Commerce is challenging the logistics industry because shipment volumes are growing rapidly and online retail typically requires more logistical work per item compared to traditional brick & mortar retail.
Online purchases require customized packing and shipping and this, on turn, requires more efficient picking processes.
Manual picking is currently preferred when the number and variety of objects increases because automation becomes increasingly challenging.
A key feature of these types of tasks is that they can be only weakly engineered (e.g. with shelf-tags for object categories; beacons for navigation) but they pose conditions that are unpredictable at design time
(objects with variable orientation-location; cluttered conditions, e.g. objects are randomly piled as in hardware stocks), a high variability of objects, and possibly the use of different robots.
Developing the technology to face these scenarios through autonomous robots is the main objective of the project .
This overall goal was pursued through the following objectives:
Objective 1: Technological realization of a demonstrator to show the potential of the autonomous learning robots to the target stakeholders.
Objective 2: Technology Assessment directed to demonstrate the technical and commercial viability of the proposed solution.
Objective 3 Business case development and product dissemination to support the implementation of an effective marketing strategy.
Online purchases require customized packing and shipping and this, on turn, requires more efficient picking processes.
Manual picking is currently preferred when the number and variety of objects increases because automation becomes increasingly challenging.
A key feature of these types of tasks is that they can be only weakly engineered (e.g. with shelf-tags for object categories; beacons for navigation) but they pose conditions that are unpredictable at design time
(objects with variable orientation-location; cluttered conditions, e.g. objects are randomly piled as in hardware stocks), a high variability of objects, and possibly the use of different robots.
Developing the technology to face these scenarios through autonomous robots is the main objective of the project .
This overall goal was pursued through the following objectives:
Objective 1: Technological realization of a demonstrator to show the potential of the autonomous learning robots to the target stakeholders.
Objective 2: Technology Assessment directed to demonstrate the technical and commercial viability of the proposed solution.
Objective 3 Business case development and product dissemination to support the implementation of an effective marketing strategy.
All objectives were fully achieved by the Project.
For Objective 1 the Partners have built a demonstrator of open-ended learning technology applied to grasping applications in a warehouse scenario.
A software architecture with multiple components was implemented first on a mockup-up form, then on a complete form able to control a robot KUKA iiiwa R800 plus a 3-finger Robotiq gripper and operate on a real scenario.
The final architecture comprises many modules which enable the robot to receive a picking order comprising multiple items, analyze the surrounding scenario to locate, identify and then pickup each item and place them into a basket.
All modules that enable grasping can be trained so that the robot can adapt to different objects and circumstances.
A dashboard enables the end user to monitor the robot global and object-specific performance so that new learning can be started when needed.
For Objective 2 the Partners have made a Technology Assessment to analyze the technical and commercial viability of the proposed solution.
In a first part of the work, warehouse stakeholders were contacted and a survey was made to derive useful KPI - key performance indicators - to be used for the Assessment.
After the demonstrator was realized, its performance was analyzed to understand the current gap and future improvements needed to bridge from the current demonstrator to a full product.
For Objective 3 the Partners have engaged into dissemination and prepared a Business Plan for the continuation of the work beyond the scope of this project, to bring the current demonstrator up to TRL9.
For the dissemination the Partners have: prepared dissemination materials, such as a website, a folder, posters and a video; ensured that the demonstrator was easy to transport and show at events; brought the demonstrator to a potential customer site for a demo and then at the Automation & Testing industry fair in Turin.
The latter event was a big success because one of the contacts at the fair immediately turned into a working collaboration for similar applications, which will help to keep developing GROW onwards.
For Objective 1 the Partners have built a demonstrator of open-ended learning technology applied to grasping applications in a warehouse scenario.
A software architecture with multiple components was implemented first on a mockup-up form, then on a complete form able to control a robot KUKA iiiwa R800 plus a 3-finger Robotiq gripper and operate on a real scenario.
The final architecture comprises many modules which enable the robot to receive a picking order comprising multiple items, analyze the surrounding scenario to locate, identify and then pickup each item and place them into a basket.
All modules that enable grasping can be trained so that the robot can adapt to different objects and circumstances.
A dashboard enables the end user to monitor the robot global and object-specific performance so that new learning can be started when needed.
For Objective 2 the Partners have made a Technology Assessment to analyze the technical and commercial viability of the proposed solution.
In a first part of the work, warehouse stakeholders were contacted and a survey was made to derive useful KPI - key performance indicators - to be used for the Assessment.
After the demonstrator was realized, its performance was analyzed to understand the current gap and future improvements needed to bridge from the current demonstrator to a full product.
For Objective 3 the Partners have engaged into dissemination and prepared a Business Plan for the continuation of the work beyond the scope of this project, to bring the current demonstrator up to TRL9.
For the dissemination the Partners have: prepared dissemination materials, such as a website, a folder, posters and a video; ensured that the demonstrator was easy to transport and show at events; brought the demonstrator to a potential customer site for a demo and then at the Automation & Testing industry fair in Turin.
The latter event was a big success because one of the contacts at the fair immediately turned into a working collaboration for similar applications, which will help to keep developing GROW onwards.
The idea of creating a versatile and adaptable picking robotic platform will make the picking processes more efficient thus resolving a number of constraints that are well known in the Logistics industry, such as:
(a) the high variability of Warehouse environments and parcels;
(b) the frequency of unpredictable manipulation errors like objects falling off the floor, or being located in the wrong place;
(c) fine-grained picking of goods, like overlapping identical objects -e.g. screws - inside a storage can;
(d) the recognition and semantic segmentation of objects characterized by different shapes, materials, rigidity (soft/hard), orientation and position on the shelf;
The Autonomous Learning capabilities of our solution would make it ideal for applications across different typologies of Warehouse.
For these reasons, our solution has a high market potential.
Order picking is the most labour intensive part of Warehouse management. Whereas automated picking is currently applied to low variety and large Stock Keeping Units (SKUs) inside engineered Warehouses, manual picking is still preferred when there is a wide variety of SKUs -for example online grocery shopping- to pick from.
Our solution goes beyond the limits of standard warehouse automation and extends the potential of automation to processes that are currently performed by human operators, with the potential of improving productivity and reducing the per-piece handling cost;
this will boost the marketability of robotics solutions in the Logistics industry and also other market sectors such as Retail with online commerce or Agrifood where a versatile picking robot could be usefully employed.
(a) the high variability of Warehouse environments and parcels;
(b) the frequency of unpredictable manipulation errors like objects falling off the floor, or being located in the wrong place;
(c) fine-grained picking of goods, like overlapping identical objects -e.g. screws - inside a storage can;
(d) the recognition and semantic segmentation of objects characterized by different shapes, materials, rigidity (soft/hard), orientation and position on the shelf;
The Autonomous Learning capabilities of our solution would make it ideal for applications across different typologies of Warehouse.
For these reasons, our solution has a high market potential.
Order picking is the most labour intensive part of Warehouse management. Whereas automated picking is currently applied to low variety and large Stock Keeping Units (SKUs) inside engineered Warehouses, manual picking is still preferred when there is a wide variety of SKUs -for example online grocery shopping- to pick from.
Our solution goes beyond the limits of standard warehouse automation and extends the potential of automation to processes that are currently performed by human operators, with the potential of improving productivity and reducing the per-piece handling cost;
this will boost the marketability of robotics solutions in the Logistics industry and also other market sectors such as Retail with online commerce or Agrifood where a versatile picking robot could be usefully employed.