Perceiving unknown and deformable objects in logistics and service robotics
Michael Suppa, Roboception GmbH Florian Töper, Mercedes-Benz AG Ashok Meenakshi Sundaram, DLRSilchersaal
Questions to be answered
The workshop will introduce the novel methods for perceiving unknown objects and focus on their applicability. The attendees will get a high-level overview on the approaches enriched with application scenarios and interaction with the speakers online and in the panel. The outcome is an assessment of the current technology in real world applications and the current limitations and risks. End-users and SMEs will get an insight into how and when to use the technology. On the other side, researchers will learn what end-users expect from the technology and how it can be successfully brought to market.
Description
Perception is one of the key technologies for enabling flexible production, such as pick and place, machine tending, kitting, assembly, and quality testing. In order to make use of machine learning and AI in these applications, the challenge for obtaining good data for seen and unseen, potentially even deformable, objects is key.
One of the key findings of the previous workshop was that the use of synthetic data in combination with real data may be one of the solutions that can address this gap. This workshop covers technology and application presentations, with a focus on the industry and end-user perspective. The technology part covers the detection of unknown objects, estimation of grasp points on unknown objects, foundation models for perception, novel object introduction and adaptation followed by real life scenarios and use cases from e.g. the RoX-project. A Slido interactive session will assess questions, user needs and expectations. A panel session with the speakers will elaborate in detail the tech trends with focus on their industrial applicability for integrators and end-users.
Organisation of the WS
14:00 Introduction and definition of key statements/questions, Michael Suppa, Roboception GmbH/University of Bremen
14:05 Perception Challenges for Kitting in Automotive Assembly Lines, Florian Töper, Mercedes-Benz AG
14:15 Insights on Pose Estimation and Grasp Prediction of Unknown Objects, Rudolph Triebel, DLR
14:25 Perceiving Deformable Linear Objects in Real-World Scenarios, Alessio Caporali, University of Bologna/Robosect srl
14:35 AI-based Perception of Seen and Unseen flexible Objects, Michael Suppa, Roboception GmbH
14:45 Interactive poll session/round table discussion with the audience
15:15 Closing remarks and take home messages
Intended outcome
The workshop will provide attendees with an in-depth understanding of AI-technology and its applicability for perceiving unknown and deformable objects. The interactive session will foster collaboration and perspectives from technology provider to end-user.
Further information
Speakers:
- Rudolph Triebel (German Aerospace Center, Germany): Insights on Pose Estimation and Grasp Prediction of Unknown Objects
- Alessio Caporali University of Bologna,Robosect srl., Italy): Perceiving Deformable Linear Objects in Real-World Scenarios
- Florian Töper (Mercedes-Benz AG, Germany): “Perception Challenges for Kitting in Automotive Assembly Lines"
- Michael Suppa (Roboception GmbH, University of Bremen, Germany):
Moderator: Michael Suppa
Panellists:
Rudolph Triebel, Florian Töper, Alessio Caporali
Organisers
Perception