国际仿生工程学会-Research Progress

More通知公告

2023-12-29

Warning: count(): Parameter must be an array or an object that implements Countable in /www/wwwroot/www.isbe-online.org_wK9SksNsZvq/libs/plugins/shared.mb_str_replace.php on line 48

Warning: implode(): Invalid arguments passed in /www/wwwroot/www.isbe-online.org_wK9SksNsZvq/libs/plugins/shared.mb_str_replace.php on line 49

仿生健康国际研讨会通知

2023-08-23

时间：2023年8月28日地点: 吉林大学（南岭校区）仿生楼209室名誉主席：任露泉，中国科学院院士大会主席：任雷（吉林大学教授），张定国（英国巴斯大学教授）共同主席：张志辉（国际仿生工程学会秘书长），韩志武（吉林大学教授）组织单位：国际仿生工程学会（ISBE）国际仿生健康联…

首页 > 阅读正文

Research Progress

On Intelligent Bionic Robot Fish

发布时间： 2018-12-18 16:28 点击：2028

The Global Summit of Artificial Intelligence and Robots (CCF-GAIR) was held in Shenzhen from June 29 to July 1, 2018. The Summit was hosted by the Chinese Computer Society (CCF), hosted by Lei Feng Net and the Chinese University of Hong Kong (Shenzhen). It was strongly guided by the government of Bao'an District. It was a top exchange event in the three fields of artificial intelligence and robotics academia, industry and investment in China. The aim is to build the most powerful cross-border communication and cooperation platform in the field of artificial intelligence in China.

CCF-GAIR 2018 continues the first two "top" teams, providing a rich platform for 1 main venue and 11 special venues (bionic robot venue, robotics industry application venue, computer vision venue, intelligent safety venue, financial technology venue, intelligent driving venue, NLP venue, AI + venue, AI chip venue, IoT venue, investor venue). It intends to provide three venues for participation. From multiple dimensions of industry, University and research, the participants present a more forward-looking and landing combination of conference content and on-site experience.

Professor Xie Guangming of Peking University Institute of Technology gave his second report entitled "Intelligent Bionic Robot Fish" at the Bionic Robot Specialty. Professor Xie Guangming gave a brief introduction to his current research. After the report, Lei Feng interviewed Professor Xie Guangming on relevant issues. Following is the interview content, Lei Feng did not change the original intention of the adjustment and editing.

Lei Feng Net: Which methods of bionic fish project need machine learning? What's the difference from the traditional method?

Xie Guangming: At present, it is mainly used in two parts. In control, it is not easy to achieve the desired control effect. Our current method is based on CFD, that is, computational fluid dynamics to enhance learning, instead of using traditional modeling and control, because its practical effect is really bad. Now we first build a mechanical model, and then let the fish learn to move in a bionic environment. For example, design a Bessel curve, then let the fish swim according to the curve, through intensive learning let the fish learn this trajectory.

On the other hand, it is the application of deep learning. In the future, there will be more and more sensors on our bionic fish. Cameras and side-line systems will collect a lot of data. We hope to use in-depth learning to get valuable information from these data, such as environmental information.

As for the difference between traditional methods, I can only say that these technologies are tools. No tool can solve all the problems. We will try different methods to solve the current problems. We have not studied the related machine learning algorithms in depth.

Lei Feng: What are the criteria for bionic target selection? Is there any specific application scenario for this project now?

Xie Guangming: We chose bionic objects, which may have been purely academic before. Now we are focusing on solving practical problems, sticking to problem-oriented and application-oriented. As for those landing on the ground, they are enterprise-oriented, civilian, like river monitoring, and underwater environment investigation. In some popular places, such as Beidaihe River, the safety of underwater environment is badly needed, but it is too heavy for people to investigate, so they can only use robotic fish to detect. The other is to test the water quality and so on. There are also individual-oriented, robotic fish to the bottom of the water to take photos and videos, entertainment is also part of the application.

Lei Fengwei: Need not biological knowledge to study bionic robots?

Xie Guangming: First of all, I think the research of bionic robots is "fetching doctrine": new materials, new energy supply methods, new sensors, new motor drivers are integrated into their own machines to improve the performance of the machine.

Second, since the study of bionic robots will inevitably involve biology, but not necessarily their own in-depth study, you can look up literature, find relevant information, such as how fish perceive the environment, you can check to obtain relevant knowledge. Of course, it's better to have suitable collaborators. If you don't have the right collaborators, you can look up the information yourself. But it's not necessary to do further research. Anyway, it's just to solve the practical problems. You can use those studies to solve the problems.

Lei Feng Net: Underwater detectors will have depth indicators. Do bionic fish have this requirement?

Xie Guangming: Depth of submergence is an engineering problem. Many solutions to engineering problems are very mature. In fact, it is the problem of processing accuracy. The reason for poor waterproofing performance may be low processing accuracy, or it may be disassembly damage caused by water leakage and so on. This is an engineering problem with mature solutions.

Lei Feng Net: What is the role of swarm intelligence of bionic fish? What are the difficulties?

Xie Guangming: For example, in the case of ocean target search, the perception and movement ability of a single robot fish are very limited, and the energy is also limited, which leads to a very small search range for a single robot fish. If robotic fish can have swarm intelligence, it will be easier to search for targets. Robot fish are autonomously controlled in the group. This kind of swarm intelligence is also called distributed intelligence. Robot fish have equal status and maintain the swarm through local interaction. In an unknown search environment, robotic fish can improve their autonomy by sharing information to improve information acquisition ability.

There are many difficulties, such as underwater communication. Individual robotic fish can only get part of the situation, and it is not clear about the latest goals. How to make information spread quickly in the population is a problem. Of course, while communicating, we should also ensure that the communication load can not be too large. There are still many problems to be solved, such as guaranteeing the interconnection of groups under the condition of limited communication.

Lei Feng Net (Public No. Lei Feng Net): What's your next research direction?

Xie Guangming: Our next research direction is to combine intelligent materials, because we found that the current means to achieve biomimetic robotic fish has reached the limit, but the effect is far from that of real fish. We are now studying the use of software materials or polymer materials, and hope to find collaborative researchers.

What's more, we also learn from the research results of some biologists on swarm intelligence, and their research is more profound. They observed that there are some simple rules in the fish flock. We want to transplant these rules to the robotic fish. We want the bionic robotic fish to have a "soul".

For example, the disorder and orderliness of fish flocks and the selfish self-protection of a single fish do not hinder the orderliness of the group are interesting phenomena. We can transplant this rule to the bionic fish to see if the bionic fish can cooperate properly. We are currently trying this deep-seated bionics.

source:https://www.leiphon