MA: Transformational Planning for Human-Robot Interaction

15.05.2009, Abschlussarbeiten, Bachelor- und Masterarbeiten

When a robot interacts with a person, it has to take the person into account when generating a plan for its course of action. This means that the robot must be aware of its own and the person's part in the plan and respect social rules for interaction. The goal of this project is to use transformational planning techniques for plan generation and to evaluate the generated plans with respect to acceptance by the human.

Project overview

We develop an autonomous household robot that adapts to its environment and interacts with a person. In this context, we have developed the transformational planning framework TRANER. Up to now, TRANER has been used to optimize plans for the robot. We have extended the plan language to represent human tasks explicitly. This extension should now be used in combination with the TRANER framework to optimize the plans for joint human-robot tasks.

Task description

• Implementation of an evaluation function for a plan based on social rules (which are available in the literature)
• Implementation of a search algorithm for finding a plan for a given situation
• Evaluation of the method with respect to efficiency and quality of plans

Prerequisites

• Basic knowledge in Artificial Intelligence
• Lisp programming skills are a plus, but not required

Kontakt: Alexandra Kirsch (kirsch@in.tum.de)

Mehr Information

http://www9.in.tum.de/teaching/student-projects/transformational-planning-for-human-robot-interaction