Towards Curious Robots: Modern Approaches for Intrinsically-Motivated Intelligent Behavior


Location: Paris, France


Curiosity is widely recognized as a fundamental mode of cognition and is particularly critical during childhood development. As such, it drives children and adults alike towards novel, previously unseen objects, movements and other stimuli, yielding new information and insights about the world and its underlying processes. Developing intelligent robots with a sense of "curiosity" may lead to an important breakthrough in artificial intelligence: agents that proactively expand their knowledge and capabilities by themselves through a snowballing process of information-generation. Recent attempts in using curiosity-inspired approaches in reinforcement learning, robot perception, task and motion planning, human-robot interaction and other fields of AI, have generated remarkable results but are largely independent of each other. Hence, there is an increasing need for a better understanding of this phenomenon at a broader scale through the collaboration of a diverse set of researchers.

In this workshop, we aim to bring together researchers, students and practitioners in robotics and other sciences to discuss how to build curious robots and agent. The workshop should appeal to both researchers developing tools to recreate aspects of intelligent agents, e.g., reinforcement learning, active learning, deep learning, symbolic reasoning, computer vision; as well as researchers in trying to understand intelligent agents, e.g., in biology, psychology, human-robot interaction. As an outcome of this workshop, we expect the participating researchers to identify and address important challenges, techniques, and benchmarks that are needed to better understand, model, synthesize and analyze curiosity-driven behavior in robots. Specifically, questionnaires will be provided to attendees in order to chart the current research activities in the field. The results of the questionnaire will be used to write a follow-up report, indicating current practices and applications as well as future challenges and opportunities. Our hope is that the resulting report could act as a manifesto for curiosity-driven learning, which can help students and researchers become familiar with the field and identify research opportunities. We also plan to award a workshop Best Poster award to encourage and recognize outstanding research and presentation efforts.