Advancing AI-Powered Robotic Cognition, Deliberation and Learning for Real-World Applications
Christian, Henkel, Bosch Research, Alessio Caporali, University of Bologna, Roberto Meattini, University of Bologna, Néstor García, Eurecat, Dimitra Triantafyllou, CERTH, Dimitrios Giakoumis, CERTH,Silchersaal
Questions to be answered
- RQ1: How can we design robotic systems that effectively reason and deliberate to expressively define and adapt their behavior in complex, dynamic environments?
- RQ2: How can robots efficiently learn new manipulation skills in unstructured environments while minimizing human intervention and maximizing adaptability?
- RQ3: How can human-centric interfaces and shared autonomy models be developed to ensure safe and intuitive collaboration between humans and robots, while leveraging prior knowledge and demonstrations?
Description
This workshop will delve into advanced methods for defining robotic behavior, with a focus on deliberation, adaptability, and real-world applications. Attendees will gain insights from leading projects, including Intelliman, CONVINCE, PILLAR-Robots, euROBIN, and MANIBOT, which are at the forefront of cognitive robotics and AI-powered systems. Key discussions will cover the essentials of robotic deliberation, examining how autonomous, context-aware decision-making can be achieved. Attendees will share input on the requirements for effective deliberation, the challenges faced, and the latest technologies enabling these capabilities.
The workshop will also explore data-efficient learning and Sim2Real transfer techniques, showcasing approaches to skill acquisition and task planning in unstructured environments. Projects will demonstrate strategies for minimizing data requirements and bridging the gap between simulated and real-world settings. Additionally, we’ll highlight the role of semantic sensor fusion and control theory to enhance robustness and situational awareness in complex tasks. Through collaborative discussions, participants will address both challenges and opportunities in cognitive robotics, aiming to shape the future of AI-powered robots for dynamic, real-world applications.
Please refer to our workshop website: https://lar-unibo.github.io/erf2025_ws42/
Organisation of the WS
Main Topic 1: High-Level Learning, Task-Level Learning, and Deliberation
- Presentations (3x9 min)
- Francisco Martín Rico (CONFIRMED)
- Title: Plan Replanning and Repair at PlanSys2
- Richard J. Duro (CONFIRMED)
- Title: The Purpose Framework: A Path to Useful Autonomous Robots
- Christian Henkel
- Title: How we to tell the robot what to do? Deliberation in CONVINCE
- Interactive Session (15 min)
- Moderated by Christian Henkel
- Online polls (Mentimeter)
- With 1 Question per speaker to collect feedback
- (e.g. Which of these technologies did you use before?)
- Audience Q&A only live from the room
- Moderator prepares 2-3 questions per speaker as backup
Main Topic 2: Skill-Level Learning and Manipulation
- Presentations (3x9 min)
- Néstor García (Inteliman project)
- Dimitrios Giakoumis (MANIBOT project)
- Title: Advancing the physical intelligence and performance of robots towards human-like objects manipulation
- João Silvério (CONFIRMED)
- Interactive Session (15 min)
- Moderated by Alessio Caporali and Néstor García
- Online polls (Mentimeter)
- With 1 Question per speaker to collect feedback
- (e.g. Which of these technologies did you use before?)
- Audience Q&A only live from the room
- Moderator prepares 2-3 questions per speaker as backup
Title: 5 - 10 words
Abstract: < 200 words
A question for the poll:
e.g. Which of these technologies did you use before?
Including answer options or free-text
Intended outcome
We don’t expect a long list of outcomes or a set of presentation slides - we want one or two results that attendees will say afterwards “We did X at ERF2025”.
- White paper (collection of presentation + summary on the discussions + any other material)
- Joint Paper
- Joint chronicle/article for website
- Special issue proposal
Further information
Organisers
None