Open up the algorithm to make the robot step more “social”

Recently, the team of Professor Quang Shuguang of the School of Psychological and Cognitive Sciences and the Institute of Brain Science and Education Innovation of East China Normal University used virtual reality combined with computational modeling methods to skillfully quantify the walking behavior of human society, and further designed an algorithm based on the characteristics of human behavior, which significantly improved the anthropomorphism and human-computer interaction experience of the robot, and effectively solved the problem of the integration of the robot into the social scene.

On November 7, the results of this interdisciplinary scientific research across humanities, sciences and engineering were published in Nature Machine Intelligence.


Photo courtesy of Kuang Shuguang

Robot social interaction puzzle

With the advent of the intelligent era, service robots are increasingly entering human life and are expected to become the right assistant of human life. How to make robots more humane, understand people better and interact with humans with high quality has become a key and difficult problem in the field of service robots.

Solving this problem requires translating theoretical concepts of human social interaction behavior into algorithms that can be implanted into robotic platforms.

In response to this problem, the research team of Kuang Shuguang successfully constructed an ideal experimental scene for measuring social behavior by using virtual reality technology, effectively quantified and modeled the walking behavior of human society, and implanted it into the navigation platform of the robot after algorithmization, realizing the three-stage leap of humanities theory from conceptualization, computing to engineering.

“Human social capabilities have irreplaceable advantages over current AI algorithms. This advantage is vividly demonstrated in the natural act of walking. “Imagine a social scenario where humans walk with humility and try to avoid interfering with the social interactions of others, although doing so may make us take a long detour and consume more time and energy.” ”

This kind of behavior and decision, which is very simple for humans, is quite challenging for robots. How to make robots have social abilities like humans, so as to improve the anthropomorphism of robots and the experience of human-computer interaction, is an important problem that needs to be solved urgently in this field.


Group photo of Kuang Shuguang (middle) team Photo courtesy of interviewee

From human social behavior to robotic navigation algorithms

“The study aims to address two important scientific and technical questions.” Zhou Chen, the first author of the paper and a postdoctoral fellow, said, “First, how to quantify and model the walking behavior of human society, and build a quantitative model to predict the walking path of humans; The second is how to implant these models into engineering algorithms to improve the robot navigation algorithm. ”

In response to the first question, although psychology, anthropology, sociology and other fields have long paid attention to human social interaction behavior, many theories have been established. However, these theories mostly stay at the level of phenomenon description and lack a complete quantitative calculation model.

The research team cleverly uses cutting-edge virtual reality technology combined with traditional experimental psychology research methods to construct an ideal social walking experimental scene in the virtual scene, and uses this experimental scene to quantitatively measure the human social interaction space and social walking characteristics. Based on the results of behavioral experiments, the team members further used computational modeling to construct a mathematical model of human walking behavior, proposed a social walking model based on individual social space, and proved the robustness of the model in complex social scenes in multiple virtual environments and real dynamic scenes.

In response to the second problem, after obtaining the computational model of social walking behavior, the research team algorithmized the model based on actual engineering requirements, implanted it into the robot platform for verification and optimization, and tested whether the algorithm based on human intelligence could improve the evaluation of robot sociality and anthropomorphism.

The results of the study prove that the computational model based on human behavior characteristics can effectively improve the experience of human-computer interaction, as well as the anthropomorphism and sociality of robots.

A new breakthrough in the intersection of arts and sciences

With the development of social and economic standards and the improvement of people’s living standards, the development of science and technology has moved from simply improving production capacity to leading the high-quality and harmonious development of human society. In such a new historical period, the importance and necessity of the interdisciplinary integration of arts and sciences have become more and more prominent.

Based on psychological theory and technology, the team of Quang Shuguang integrates virtual reality, computer science and other fields, and builds a multidisciplinary cross-integration research team, which is committed to using the latest intelligent technology to quantify a series of abstract and conceptual humanities theories, and using humanities ideas to stimulate and guide breakthroughs in the field of science and technology.

In recent years, the research team has designed a series of new research paradigms to realize the calculation and engineering of humanities theory, and the relevant results have been published in international high-level journals in the fields of psychology, neuroscience, and computer science, becoming a unique research team in the field of interdisciplinary research of arts and sciences.

“This is an interdisciplinary science study that spans the humanities, sciences and engineering.” “This achievement constructs the connection between human behavior theory and navigation algorithm, and realizes the leap of human social walking theory from concept, model to algorithm.” (Source: China Science News, Zhang Shuanghu, Huang Xin)

Related paper information:

Source link

Related Articles

Leave a Reply

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

Back to top button