A deep learning model that generates the nonverbal social behavior of robots

A deep learning model that generates the nonverbal social behavior of robots

Generating bot social behavior involves mapping the next bot behavior in order to respond to the user’s current behavior while maintaining continuity with the bot’s current behavior.

Researchers at the Electronics and Communications Research Institute (ETRI) in Korea recently developed a deep learning-based model that can help produce engaging nonverbal social behaviors, such as hugging or shaking someone’s hand, in robots. Their model, presented in a previously published paper on arXiv, can actively learn new, context-appropriate social behaviors by observing interactions between humans.

Deep learning techniques have yielded interesting results in areas such as computer vision and Understanding natural languageWoo-Ri Ko, one of the researchers who conducted the study, told TechXplore. We started applying Deep learning to me social robotsspecifically by allowing the bots to learn social behavior of human-human interactions on their own. Our method does not require prior knowledge of human behavior models, which are usually costly and time-consuming to implement.”

The artificial nerve network (ANN) – The architecture developed by Kuo and colleagues combine Seq2Seq (sequence-to-sequence) model introduced by Google researchers in 2014 with generative adversarial networks (GANs). The new architecture was trained on AIR-Act2Act dataseta collection of 5,000 human-to-human interactions that occur in 10 different scenarios.

“The proposed neural network architecture consists of an encoder, a decoder, and a discriminator,” Ku explained. “The encoder encodes the current user behavior, the decoder generates the next bot behavior according to the behavior of the user and the current bot, and the discriminator prevents the decoder from outputting invalid mode sequences when generating a long-running behavior.”

5000 interactions included in AIR-Act2Act dataset They were used to extract more than 110,000 training samples (that is, short video clips), in which humans performed specific nonverbal social behaviors while interacting with others. The researchers specifically trained their model to generate five nonverbal behaviors for the robots, namely bowing, staring, shaking hands, hugging, and blocking their faces.

Kuo and colleagues evaluated their model for generating nonverbal social behavior in a series of simulations, specifically applying it to a simulated version of Pepper, a A robotic humanoid creature which are widely used in search settings. Their initial results were promising, as their model successfully generated the five behaviors that were trained at appropriate times during simulated interactions with humans.

“We have shown that it is possible to teach robots different types of social behaviors using a deep learning approach,” Kuo said. “Our model can also generate more natural behaviors, rather than repeating predefined behaviors in the rules-based approach. With the robot generating these social behaviors, users will feel that their behavior is understood and emotionally cared for.”

The new model created by this team of researchers could help make social bots more socially adaptive and responsive, which could in turn improve the overall quality and flow of their interactions with human users. In the future, it could be implemented and tested on a wide variety of robotic systems, including home service robots, routing robots, delivery robots, educational robots, and telepresence robots.

“We now intend to conduct further experiments to test the robot’s ability to exhibit appropriate social behaviors when deployed in the practical world and human encounter; the proposed behavior generator will be tested for its robustness with the noisy input data that the KO robot is likely to acquire.” Furthermore, by collecting and learning more interaction data, we plan to expand the number of social behaviors and complex actions that a Robot can appear.”

more information:
Woo-Ri Ko et al, Generation of Nonverbal Social Behavior of Social Robots Using Inclusive Learning, arXiv (2022). DOI: 10.48550/arxiv.2211.00930

Ilya Sutskever et al, Sequence to Sequence Learning with Neural Networks, arXiv (2014). DOI: 10.48550/arxiv.1409.3215

Woo-Ri Ko et al, AIR-Act2Act: A human-to-human interaction dataset for teaching nonverbal social behaviors to robots, Available here. International Journal of Robotics Research (2021). DOI: 10.1177 / 0278364921990671

© 2022 Science X Network

the quote: Deep Learning Model Generates Nonverbal Social Behavior of Robots (2022, November 24) Retrieved November 24, 2022 from https://techxplore.com/news/2022-11-deep-generates-nonverbal-social-behavior.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.


#deep #learning #model #generates #nonverbal #social #behavior #robots

Leave a Comment

Your email address will not be published. Required fields are marked *