Abstract des Autors

The objective of the interdisciplinary Collaborative Research Center 588 “Humanoid Robots – Learning and Cooperating Multimodal Robots” is the construction of a humanoid robot. To make the robot attractive for services in households and offices the robot should be able to communicate and to interact with humans. Among other topics, this humanization concerns the strategies of motion planning for the humanoid robot. Besides anthropometric resemblance, the motion of the robot must be comparable to human motion. In this work, we analyze trajectories of limbs and develop guidelines for motion planning based on task-specific characteristics. Based on these guidelines we present adaptive models that are trained to classify single motion phases. A new algorithm is presented to compose these elementary models to large models that recognize complete motion trajectories. Finally, we show how to use the motion models to control the movements of a humanoid robot. Verf.-Referat