Overview. This Dagstuhl Seminar will study the application of modern Knowledge Representation methods in robotics. An interdisciplinary discussion to establish a roadmap for future collaborations will be held between researchers and practitioners of different domains – Knowledge Graph experts to help applying neuro-symbolic technologies on structured knowledge sources, robotics experts that provide knowledge on robotic infrastructures and task deployment, and Hybrid Intelligence experts to help designing scenarios where robots and humans collaborate in a synergistic, proactive and trustworthy way.

Motivation. AI and Robotics technologies affect our daily lives at an ever-increasing pace. Vacuum cleaners with advanced navigation abilities are verbally instructed to “clean the carpet”, hospital robots manipulate surgical tools with Deep Learning-based vision, companion service robots use language models for affective communication with patients. These robots will soon be expected to collaborate synergistically and proactively with humans toward shared goals, enhancing each other’s capabilities toward a new partnership recently defined as Hybrid Intelligence (HI).

To achieve complex tasks in real-world situations, robots need to deal with large sources of multimodal knowledge representing themselves, the others, and the environment around them. The Knowledge Representation (KR) community has been studying techniques to manage (model, reason, and engineer) heterogeneous and large-scale knowledge sources for a long time. Modern Knowledge Representation techniques, i.e. neuro-symbolic methods, semantic technologies, and Knowledge Engineering at scale, allow nowadays to handle large-scale heterogeneous structured knowledge (Knowledge Graphs). The application of these modern KR techniques is an opportunity for robots to tackle long-standing challenges, including acquiring/identifying relevant knowledge for a task (including long-term ones), aligning tasks and goal representations, and collaborating with humans in a trustworthy way.

Topics and Goals.The Dagstuhl Seminar will be focused on three areas addressing three core challenges: (1) Ontologies to enable Robotic Interoperability and Knowledge Transfer, (2) Knowledge Engineering to foster Human-centeredness during a hybrid human-robot collaboration, and (3) Neuro-symbolic solutions to allow Explainable Agency. Each area will be analyzed according to the three types of KR methods: representation (i.e. modelling choices), reasoning (i.e. inference abilities), engineering (i.e. implementation solutions), to determine existing solutions, limitations, and future directions. Besides determining the field’s state-of-the-art (from modelling, inference, and engineering solutions to related resources and support tools), the seminar participants will identify a list of challenges and new research opportunities to be addressed by the community as a future roadmap. Ultimately, we will work on the next steps to set up an interdisciplinary network to jointly work on these challenges and exchange ideas.

Structure and Outcome. The Dagstuhl Seminar will include lightning talks, invited keynotes, and breakout sessions to mix frontal discussions and more active brainstorming. The seminar will also aim to prepare a final report summarizing the discussions and roadmap as a journal position paper/manifesto to be shared with the wider community.

Creative Commons BY 4.0

Ilaria Tiddi, Alberto Olivares Alarcos, Philipp Cimiano, and Michael Beetz