Skip to main content

Obtaining good data for Agile production, logistics and lab automation

Track: Robotics and AI / Data

Date and Time: 13 Mar 2024 14:40-16:00

Organiser: Michael Suppa

Main questions to be answered: Model data of the final product is usually available in agile production. How do you deal with the potential lack of knowledge during the production process when object handling is required at intermediate steps?

Additional questions: Synthetic training data generation requires model data for simulation. Do you have this data available, where does it come from and how do you asses the level of correctness?

Additional questions 2: Which level of expertise regarding 3d vision and machine requirements are available in your area? (none, beginner, moderate, expert)

Workshop Description: Perception is one of the key technologies for enabling flexible production, such as pick and place, machine tending, assembly, and quality testing. In order to make use of machine learning and AI in these applications, the question for obtaining good data is becoming more and more relevant. Especially in flexible automation, models of the part are usually available. However, they may represent the final product and not the product as seen in the current production stage. In warehouse logistics and lab automation, the model data if available at all, may be even less reliable. One of the key findings of previous workshop was the use of synthetic data in combination with real data may be one of the solutions that can address this gap. This workshop will be dedicated to answering the key question on how to address model inaccuracies in new ways to truly enable flexible production. As an outcome, the use and access to good data for three domains will be elaborated to enable end-users, integrators and perception system suppliers to fully exploit and understand the potential of mixed data approaches.

Intended Outcome: Workshops engaging non-roboticists to understand needs/goals (policy, commercial, technical), Workshop on recent developments in technology or applications, Workshop discussion topics of common interest, success stories, use cases, etc, Workshop covering material new to ERF that the Robotics community should engage with

Approach: 00:00 Introduction and definition of key statements/questions, Dr. Michael Suppa, CEO, Roboception GmbH 00:10 Towards Tricky (Transparent, Reflective, …) Object Models and Detection, Prof. Markus Vincze, Professor, TU Vienna, Austria 00:20 AI-driven perception logistics, Dr. Radhita Gudipati, Ocado Technologies, UK. 00:30 Data generation for Lab Automation, Dr. Patrick Courtney, CEO, Tec-connection, UK 00:40 The power of synthetic data in Agile Production, Michael Suppa, CEO, Roboception GmbH, Germany 00:50 Interactive poll session/round table discussion with the audience 01:15 Conclusion and take home messages

Contributors: Prof. Markus Vincze, Professor, TU Vienna, Austria, confirmed, Role: Research and AI perspective Dr. Radhita Gudipati, Ocado Technologies, UK, confirmed, Role: warehouse automation perspective Dr. Patrick Courtney, CEO, Tec-connection, UK, confirmed, Role: lab automation perspective, confirmed Dr. Michael Suppa, CEO, Roboception GmbH, Germany, confirmed, Role: Agile production perspective and representative of TG perception

Further Information: Link to the website of the previous workshop: https://roboception.com/en/innovation-en/erf2023/