Humans can effortlessly complete a wide range of challenging tasks in our unstructured and uncertain real world. However, robots still have a long way to go before they develop “embodied” artificial intelligence. Thanks to powerful neural models and extensive data sets, robot learning has made remarkable progress. Nevertheless, there are still open research questions: Are massive architectures and data required to endow robots with “artificial embodied intelligence” to learn to solve tasks that are intuitive for humans? And how can we make substantial progress toward robust and adaptive robot learning systems that operate reliably in the dynamic real world?

The SIREN (“Structured Interactive Perception and Learning for Holistic Robotic Embodied Intelligence”) project by Georgia Chalvatzaki is dedicated to answering these questions. As part of her ERC-funded project, the researcher will investigate the principles that underlie the complex interaction between robots and the environment and study how this interaction evolves. SIREN proposes a new systemic view of how robots can learn: a unique, comprehensive representation of the robot and its environment as a unified system. Through this view, the project will research what features of the perception–action cycle are crucial for the development of robotic embodied intelligence.

The results of SIREN will enable robots such as humanoid mobile manipulators – robotic systems with arms, heads, and a mobile base – to autonomously find their way around in unstructured human environments and complete challenging tasks that require the smooth and efficient coordination of perception and action. The paradigm shift proposed by SIREN opens avenues for future groundbreaking research rooted in SIREN’s impacts toward continuous robot learning systems that are integrated into and evolve with their environment.

sip