Visual ObservationforHybrid Intelligent
Previous: Observer Construction
We utilize the observer framework to recognize error states and sequences. The idea behind this recognition task is to be able to report on visually incorrect sequences. In particular, if there is a predetermined observer model of a particular task under observation, then it would be useful to determine if something went wrong with the manipulation actions. The goal of this reporting procedure could be to alert operators or possibility to supply feedback to the manipulating robot so that it could correct its actions. Some examples of errors in manipulation include unexpected behaviour of the system, such as objects falling unexpectedly from the manipulating hand during a grasp and lift operation or some visual errors like unexpected occlusions between the observer camera and the manipulation environment.
There are a number of ways in which these problems could be reported. One such way can be to comply with the navigation strategy that was described in [8] in order to capture the current state (i.e. incrementally update belief in the current state and sequence as event matches occur while ``navigating'' the automata). If no match occurs, then the error would have to be reported. The correct sequences of automata state transitions can be formulated as the set of strings that are acceptable by the observer automaton. This set of strings represents precisely the language describing all possible visual task evolution steps.