In this paper, a multi-level approach to intention, activity, and motion recognition for a humanoid robot is proposed. Our system processes images from a monocular camera and combines this information with domain knowledge. The recognition works on-line and in real-time, it is independent of the test person, but limited to predefined view-points. Main contributions of this paper are the extensible, multi-level modeling of the robot's vision system, the efficient activity and motion recognition, and the asynchronous information fusion based on generic processing of mid-level recognition results. The complementarity of the activity and motion recognition renders the approach robust against misclassifications. Experimental results on a real-world data set of complex kitchen tasks, e.g., Prepare Cereals or Lay Table, prove the performance and robustness of the multi-level recognition approach.

Estimating a user's intention is central to close human-robot cooperation. In this paper, the problem of performing intention recognition with tree-structured Dynamic Bayesian Networks for large environments with many features is addressed. The proposed approach reduces the computational complexity of inference O(bs) for tree-structured measurement models with an average branching factor b and tree height s to O(b̃s), where b̃ ≪ b. The key idea is to switch between a finite set of reduced system and measurement models in order to restrict inference to the most important features. A model predictive approach to online switching between the reduced models is proposed that exploits an upper bound of the distances of the reduced models to the full model. The effectiveness of the proposed algorithm is validated in the intention recognition for a humanoid robot using a telepresent household scenario.