FLEXIBLE DIALOGUE BETWEEN HUMAN AND ROBOT USING REINFORCEMENT LEARNING

• Main Author: Rofi'ah, Binashir
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49317
• Institution: Institut Teknologi Bandung

Dialogue has been widely used for verbal communication between human and robot interaction, such as an assistant robot in hospital. However, this robot was usually limited by a predetermined dialogue, so it will be difficult to understand new words for a new desired goal. In this paper, we discussed conversation on entertainment, motivation, emergency, and helping with knowledge growing method. Natural language processing was used to search unique words, and process new words that robot didn’t understand. The new words were called new knowledge, this knowledge was then added to database/initial Q-table. After that, Reinforcement Learning will learn to get relation between old database and new knowledge, the methods used were Q-Learning and SARSA. To cheer up patient, we provided playing mp3 audio for music, fairytale, comedy request, and motivation. The execution time for this request was 3.74 ms on average. In emergency situation, patient able to ask robot to call the nurse. Robot will record complaint of pain and inform nurse. From 7 emergency situation report, robot able to save complaint on database equipped with time recording. In helping conversation, robot will walk to pick up belongings of patient. Once the robot didn’t understand with patient’s conversation, robot will ask until it understands. From asking conversation, knowledge expanded from 2 until 10 with learning execution from 1500 ms until 3500 ms. Q-Learning takes longer time than SARSA, however, both were achieved desired object in 200 episodes. It concludes that method to overcome robot knowledge limitation in achieving new dialogue goal for patient assistant were formulated.