Robotar med kausala förmågor

Causal reasoning is crucial for development of truly intelligent robots, in particular for robots working in close contact with humans. However, published robotics research rarely addresses causality explicitly.

The overall aim of theproposed project is therefore to investigate how causal reasoning may be introduced and utilized in human-robot interaction (HRI).To reach the aim we will develop methodology for causal learning, by which a robot infers causal relations based on observations in combination with information from dialogue between the robot and interacting humans.

In this way the learning process benefits from the superior causal reasoning of humans.

As another part of the project, we will consider the consequences of causal reasoning robots for interacting humans.The project is planned for a 4-year period, and the applied funding will mainly finance a 2-year postdoc and support the work by the multidisciplinary project team comprising the applicant who specializes in robotics and machine learning, and two participating researchers specializing in statistical causal learning and natural language processing.

The planned research will be a mixture of algorithm development and implementations and user studies on physical robots.