Performance evaluation of visual servoing based on complete feature sets

Zhe Ma, Jian-Bo Su

Abstract: In terms of visual feedback, the process of visual servoing could be characterized by the changes of the visual features. The concept of feature completeness for visual servoing is introduced. If there exists a feature set that varies from the initial state to the target state while the prescribed task is finished, then the feature set is complete for this task. With such a feature set, the visual servoing task is transformed to designing the control algorithm to alter every feature to the corresponding final state optimally. Visual servoing systems are firstly categorized in terms of feature completeness. Then how feature completeness affects the performance of task fulfillment is discussed. The strategy of visual servoing control is investigated based on complete feature sets, incomplete feature sets as well as redundant feature sets, respectively. Finally, simulation and corresponding results are described to demonstrate how feature completeness functions and the advantage of this technique.