BEHAVIOR LEARNING METHOD OF ROBOT

• Main Authors: LIM, JONG RAK, KIM, NAM UNG, JANG, SEONG JUN
• Format: Patent
• Language: English; Korean
• Published: 12.05.2003
• Edition: 7
• Subjects: CONTROLLING; PHYSICS; REGULATING; SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES

PURPOSE: A method for learning a behavior of a robot is provided to obtain a natural leaning effect by learning the frame information related to voice commands. CONSTITUTION: A word database includes words for performing particular behaviors. Accordingly, the words in the word database are related to the particular behaviors. The word which is not related to the particular behavior is recognized in a word recognition process(SP1). A behavior study mode is performed according to an input state of the word related to the particular behavior of the word database in order to perform a stopping operation and a recording operation(SP2). A variation of a stance of a robot is checked within a predetermined time(SP3). An autonomic mode is performed if the stance of the robot is not varied within the predetermined time(SP4). The current stance is recorded as a final frame and operation information is generated by interpolating the initial frame and the final frame if the varied stance is maintained(SP5,SP6). 본 발명은 로봇의 행동 학습방법에 관한 것으로, 로봇의 정지자세를 로봇이 가지고 있는 모든 관절의 현재 지시값을 프레임 정보로 정의한후,그 프레임 정보와 음성명령을 연결시켜 학습시키도록 한 것이다. 이를 위하여 본 발명은 음성DB중 행동과 연관되지 않는 단어가 음성으로 인식되면 행동학습 모드로 전환하는 제1 과정과; 사용자에 의해 변화된 자세를 상기 단어에 해당되는 동작정보로 생성하여 학습시킨후, 자율 모드로 복귀하는 제2 과정으로 수행한다.