Selected papers | AI learning must also set a small target, so that artificial intelligence has social awareness - "socially conscious multi-agent learning: the best solution for the community"

Application Scenario Guidance: Multi-agent systems, such as mobile sensors, unmanned vehicles, etc., are an important branch of distributed artificial intelligence research. Because of their strong fault tolerance, robustness, and scalability, they are Widely used in military, industrial and agricultural production, medicine, transportation, services and other fields. A multi-agent system is a collection of multiple agents with computing and mobility capabilities. Each agent is a physical or abstract entity that can act on itself and the environment and communicate with other agents.

title:

Socially aware multi-agent learning: an optimal social solution

Summary:

In a multi-agent system, learning ability is crucial to each agent, which is related to how it normally responds to unknown opponents in a dynamic environment. From the system designer's point of view, it is highly desirable that agents learn to collaborate with socially optimal solutions while avoiding being exploited by selfish opponents. For this reason, we propose a novel gradient elevating algorithm (SA-IGA), which enhances the basic gradient elevating algorithm by incorporating social awareness into the strategy update process. We theoretically analyze the learning dynamics of SA-IGA based on dynamic system theory, and SA-IGA has linear dynamics in many games including symmetric games. A detailed analysis of the learning dynamics of the two representative games ("Prisoner's Dilemma" games and coordination games) was conducted. Based on the SA-IGA concept, we further propose a multi-agent learning algorithm based on Q learning update rules, called SA-PGA. The simulation results show that the SA-PGA agent can obtain higher social welfare than the previous social optimal rule for conditional joint action learners (CJAL) and is resistant to independent rational opponents through Nash equilibrium solutions. .


The first author introduction:

Li Xiaohong, female, born in September 1965, Ph.D. in Engineering, Department of Computer and Information Technology, Professor and Doctoral Supervisor at Tianjin University. Senior member of Computer Society, ACM member, member of Software Engineering Committee; Executive Director of National Computer Education Research Association of Colleges and Universities; Female Worker of Tianjin University, Vice-Chairman of the College Trade Union. In recent years, he has devoted himself to the research of security software engineering, trusted software, and information security.

Senior Member of Computer Society, Member of ACM, Member of Software Engineering Committee; Executive Director of Computer Education Research Association of National Colleges of Higher Education; Expert of National Science and Technology Awards Review; Review of Degree Thesis of Graduate Education and Development Center of Ministry of Education, Subject Construction and Evaluation Consultant; Reviewers of Journal of Computer Science, Computer Science and other magazines; International Journal of Software Engineering and Information Security; International Conference Review Expert; Member of Academic Degree Committee of Computer College, Academic Committee Member, Academic Development Committee Member, Teaching Steering Committee Member Concurrent secretary; member and secretary of the College's 985 expert group; female worker member of Tianjin University, vice chairman of the college's labor union.

In recent years, he has presided over or participated in the completion of nearly 20 national, provincial, and horizontal scientific research projects. In recent years, he has published more than 50 academic papers, including more than 20 major diplomas for degree and postgraduate education, more than 20 international conferences, and EI search. More than 20 articles, SCI search 6 articles. More than 20 national invention patents have been declared and 6 have been authorized. Software copyright 4. 1 monograph, 1 provincial and ministerial science and technology award. As the project leader, he is responsible for hosting one national key project (subproject leader), one national fund project, one basic Tianjin project, and one corporate cooperation project.


Via PRICAI 2016

Paper original file download

Lei Feng Network Press: This article by Lei Feng network (search "Lei Feng network" public number concern) exclusive compilation, without permission prohibited reprint!

Plunger Seat Safety Switch

YESWITCH FD01 Snap Mount Plunger Safety Seat Switch

YESWITCH FD01 plunger snap mount safety seat momentary switches sense operator presence. With Normally Open Normally Open function,used in conjunction with a machine or vehicle`s safety system, operator presence switches provide a fail-safe condition (shutting down the engine, turning off cutting blades, etc.) when an equipment operator is no longer in the seat. Sealed, unsealed and high-cycle-life Hall Effect styles are available. Choose from twist lock or bolt mounting. With decades of proven reliability, our double pole push button switches are ideal for lawn mowers,golf cart and electric-powered applications.

SPECIFICATIONS
RATING: DC12V,3A
CONTACT RESISTANCE: 100m MAX.
INSULATION RESISTANCE: 1000M MIN.
DIELECTRIC STRENGTH: 1000Vrms min.
OPERATING FORCE: 2.0Kgf
OPERATING TEMP: -40~+85 Degree
MECHANICAL LIFE: 75,000 CYCLES

MATERIALS
HOUSING: Nylon,Black.
PLUNGER: PBT, Gray.
CASE: Nylon,Gray

TYPICAL APPLICATIONS INCLUDE
Outdoor Power Equipment,Utility Vehicles,Power Sports,Construction,Farm & Agricultural,Asphalt Pavers/Concrete Mixes,Light Industrial,Material Handling
Off-Highway Vehicles,Mining

Fd01 03

Interlock Plunger Safety Switch,Riding Mower Safety Switch,Mower Plunger Safety Sswitch

YESWITCH ELECTRONICS CO., LTD. , https://www.yeswitches.com