The successful operation of any autonomous agent heavily depends on capturing and understanding the complex nature of the environment, including its volatile or implicit characteristics.
Even though robotics is undergoing rapid development, several application areas are still hindered by limitations in how environmental information is represented. Qualities that might be missing from state-of-the-art representations include the relations between environment features, meta-information describing the quality and applicability of the representation itself, or condition-dependent characteristics. In this workshop, we go beyond a discussion of traditional 'semantic maps' (simply with object annotations), and instead gather experts from a range of fields to highlight and address limitations in conventional spatial representations for robotics. We bring together researchers with diverse and versatile expertise and backgrounds to achieve this goal. Through gathering such a group, we aim to fulfill the following three objectives:
- We aim to provide a platform to formulate and communicate the most urgent needs for representations in robotics.
- We want to expose the robotic community to recent developments in relevant correlated fields such as AI and cognitive science.
- As a result, we want to pave the road for developing novel representations reaching beyond the state-of-the-art.