Abstract: Traditional methods of teaching programming cannot discriminate the effective operations from misoperations, which are inefficient and sometimes invalid as applied to service robots. A method of teaching programming is proposed to solve this problem, which can extract effective operations from the demonstration track online and integrate them as well as reappearance. Demonstration track is divided into segments based on improved mean squared velocity algorithm, so as to convert the original track into bunches composed of segmented tracks. Various segmented tracks are identified via nearest neighbor rule to achieve feature extraction. Fault-tolerant string matching algorithm is applied to find the characteristics of tasks and then achieves the effective part extraction of demonstration trajectory from bunches. Typical experiments grasping a cup and putting it down on the specified location from different positions are completed via an intelligent wheelchair mounted manipulators, and the results well verify the effectiveness of the teaching method based on effective operation feature extraction.

Key words: Effective operation feature, Mean squared velocity algorithm, Pattern recognition, Service robot-arm, Teaching programming