We described a system for observing a manipulation process. The proposed approach can be generalized for other hybrid systems involving different kinds of quantization requirements for dynamic systems, for sets of discrete, continuous and symbolic parameters. The use of discrete event dynamic systems with uncertainty modeling for the event description enables the observer to recognize tasks robustly. The proposed system also utilizes the a-priori knowledge about the task domain in order to achieve efficiency and practicality. The high level formulation allows for recognizing and reporting on the visual system state as a symbolic description of the observed tasks.
Thus, we have proposed a new approach to solving the problem of observing a moving agent. Our approach uses the formulation of discrete event dynamic systems as a high-level model for the framework of evolution of the visual relationship over time. The proposed formulation can be extended to accommodate for more manipulation processes. Increasing the number of states and expanding the events set would allow for a variety of manipulating actions.