Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
This thesis makes a contribution to autonomous robotic manipulation. The core is a novel constraint-based representation of manipulation tasks suitable for flexible online motion planning. Interactive learning from natural human demonstrations is combined with parallelized optimization to enable efficient learning of complex manipulation tasks with limited training data. Prior planning results are encoded automatically into the model to reduce planning time and solve the correspondence problem.