Information-driven multi-robot behavior adaptation to intention in human-robot interaction

A multi-robot behavior adaptation mechanism to human intention is proposed in human-robot interaction, where the information-driven fuzzy friend-Q learning (IDFFQ) is used to generate an optimal behavior-selection policy, and intention understanding is done mainly based on human emotions. It aims to endow robots with the human-oriented interaction capabilities in understanding and adapting their behaviors to human intentions, which decreases the response time of robots by embedding the humans' identification information (e.g., religion) for behavior selection, and increases the satisfaction of humans by considering humans' deep-level information (e.g., intention and emotion), so as to make interactions run smoothly. Experiments are performed in a scenario of drinking at a bar. Results show that the learning steps of the proposal is 51 steps less than that of the fuzzy production rule based friend-Q learning (FPRFQ), and report's robot time (computational time) is about 1/4 of the time consumed in FPRFQ. Additionally, emotion recognition and intention understanding receive an accuracy of 80.4% and 85.71%, respectively. Moreover, subjective evaluation of customers through questionnaire obtains a satisfaction of "satisfied". Based on the preliminary experiments, the proposal is being extended to service robots for behavior adaptation to customers' intention to drink at a bar.