Photo: ABB

Today, industrial automation is able to solve very complex tasks in controlled environments. Due to time-consuming and high knowledge demands for setup, its use is however limited to large scale production. In addition, the low maturity level of artificial intelligence in safety critical automation, limited availability of data and the need of explainability and highest possible reliability, much research effort is needed on the way towards broad scale robotics and machine learning for Industry 4.0. 

WARA Robotics supports research connected to mainly two types of common industrial tasks: kitting and assembly. Kitting is the process in which all material needed for assembling a product is collected and delivered to a workstation. This task requires refined navigation and manipulation skills such as pick and place. Assembly is the task performed when the pieces are put together. This demands contact-rich manipulation where the contact dynamics reflecting the physical properties of the pieces exhibit a high degree of variation. The two tasks exhibit a wide range of research challenges crucial for the further development of industrial automation.

During 2021, WARA Robotics will construct a physical testbed at ABB Västerås including robots, material and other resources needed to deploy research in a multitude of aspects related to kitting and assembly. The setup will allow benchmarking of results produced by different research groups. Read more under Resources and Services further down.

Objectives for participation

Research under realistic scenarios is important for two reasons; to achieve higher technology readiness level, tests and demonstrations in industrial environments are required, and exploring technologies in their relevant environments allows for discovery of unexpected challenges otherwise invisible.

Research

• Access to testbeds in an industrial setting
• More relevant and new research questions
• Aid visibility, practical relevance, and impact of WASP research 

Education

• Support the WASP Project Course
• Increased network facilitates a future research career in industry or academia

Industry

• Build new and strengthen existing networks between industry and academia
• Faster knowledge transfer between academia and industry

Research Focus Areas

Possible research areas connected to kitting and assembly that can be investigated in the WARA Robotics are as follows:

Reinforcement Learning

Reduction of samples

Sim-to-real transfer

Multiple robot data generation

Automated Reasoning

Task planning

Resource scheduling

Motion planning

Multi-robot coordination

Perception

Segmentation & scene understanding

Object recognition & tracking

Multi-modal perception

Human-Robot Interaction

Multi-modal interaction

Learning from demonstration/observation

Assembly with Dual-arm Robots

Utilization of dual arms in assembly

Resources and Services

Hardware setup

• Several robot arms on mobile platforms equipped with cameras or laser scanners.
• Learning-by-demonstration motion capture system.
• Hardware for accurate ground-truth positioning of robots, tools and other objects.
• A room with several tables or stations in which the robot will move.
• Warehouse mock-up, reflecting the variety of a typical warehouse particularly for picking:
  ○ variety of parts
  ○ variety of containers
  ○ variety of shelves

• Various assembly challenges from different industrial collaborators
• Fixtures and rigs needed for specific assembly operations
• 3D-printing functionality
• 2-3 stationary single arm/dual arm robots for the multi-robot skill learning station
• Sensors: force, haptic, vision, LiDAR, 3D-vision, laser-scanners, pressure and proximity
• Stationary computers in the lab for running the experiments, ready with all necessary software installed as listed in the SW set-up below. Necessary computing resource for data management and vision data
• Wireless communication systems, both Wi-Fi and 5G

Software setup

• One or more simulation environments for development and executing learning algorithms (such as Algoryx).
• Simulation environment where it is possible to include sensors such as LiDAR, radar and camera in combination with simulation of robots and mobile platforms.
• Modular ROS-based base implementation of task planning, motion planning, multi agent path finding for both mobile base and dual arm robot for easy benchmarking and improvement of modules.
• ROS-based software interfaces for communication between PC, robot and other equipment.
• Data-driven applications can be linked to a cluster for the computation (WARA Common) and possibly to the ABB Ability platform.
• On case-by-case basis and with separate two-part agreement:
  ○ Low level interfaces to control the robots (using EGM-RI or equivalent), where it is also possible to synchronize the low-level control between mobile platform and robot.
  ○ Robot models, kinematic as well as dynamic models, for control and optimization.

• Documentation/guidelines of the above mentioned

The Core Team

Jonas Larsson, Project Manager WARA Robotics, ABB Corporate Research

Joachim Sachs, Ericsson

Kenneth Bodin, Algoryx

Academic partners TBA.