Towards Robust Human-Robot Interaction - A Quality Diversity Approach

This course aims to teach learners about the importance of automatic scenario generation in human-robot interaction using a quality diversity approach. The learning outcomes include understanding how quality diversity algorithms can help discover failure scenarios, improving search in scenario space, and generating complex and realistic scenarios. The course covers topics such as MapElites, diversity measures, space distortion, failure cases, gradients, and applications in procedural content generation and human preference learning. The teaching method involves a seminar-style talk by an expert in the field. This course is intended for students and professionals interested in robotics, human-robot interaction, and autonomous systems.