Human-robot interaction

Robots interact with people in ever more profound ways as they are being used in domestic and public environments. These environments are typically unknown, dynamically changing and populated by people. In the civil domain robots appear as (museum) tour guides, or travel agents. In health care robots support independent living or assist care personnel. In all these applications it is tacitly assumed that robots possess the cognitive intelligence to perform these tasks, but this is often far from being true. To interact with an individual a robot needs to approach a person, attract and monitor attention, and possess context awareness. In multi-party settings turn taking and joint attention are fundamental cognitive skills for a robot. For natural HRI it would seem useful if a robot could understand and provide social cues like co-speech gestures, and facial expressions. Should a robot be persuasive, entertaining or submissive? Do robots need a theory of mind?

The course Human-Robot Interaction addresses some of the fundamental problems of interacting with humans. It combines knowledge and experience from cognitive sciences, artificial intelligence and robotics. The course starts with explaining how a probabilistic framework can be used to incorporate context information from noisy sensors in the robot’s world model. The next step is to make robots person aware by recognizing human behaviour and by providing recognizable behaviour. Finally, probabilistic reasoning and decision making is added to enable autonomous cognitive models of human-robot interactive behaviours. As part of the course, students implement their cognitive models for a given context on a robot and investigate the requirements for social intelligence of robots.

Assumed preknowledge:

It is strongly advised that students refresh their knowledge of probability theory, integration and differentiation of functions of multiple variables. In particular, proficient knowledge of the following topics before starting this course is strongly recommended: Discrete and continuous probability distributions; Expected value, variance and covariance; Conditional probability and Joint probability; Law of total probability, and Bayes rule.

For Linear Algebra prior knowledge (like 2DE20 Mathematics 1):; Functions of multiple variables; Matrix algebra; Vector spaces.

(For Python programming only basic programming knowledge is required; This includes basic syntax, variables, conditional statements and loops, SDKs and other software tools for accessing the robots are provided and explained in class).

After this course, students are able to:

• Develop biologically inspired probabilistic models for human-robot interaction by applying concepts such as Bayes filters, Kalman filters, and particle filters.
• Analyze state-of-the-art technologies for human-robot interaction, focusing on perception, decision-making, and communication systems.
• Apply experimental methodologies to validate robot performance and user experience, focusing on social interaction metrics such as gesture recognition, turn-taking, and eye-contact.
• Implement models on humanoid robots, including Nao, Pepper, or Misty, to enable natural and socially-aware interactions.

You must meet one of the following collections of requirements

• Collection 1
• Completed Final examination Bsc program
• Collection 2
• Completed Pre-Master

• Selected articles. (handout)
• Sebastian Thrun, Wolfram Burgard, and Dieter Fox. Probabilistic Robotics. MIT Press, 2005