Shulou(Shulou.com)06/03 Report--

This article mainly explains "analyzing the technology and application of swarm intelligence for shared travel". The content of the explanation is simple and clear, and it is easy to learn and understand. let's study and learn "analyze the technology and application of swarm intelligence for shared travel".

The recent breakthroughs in artificial intelligence technology mainly focus on effectively simulating individual intelligence, but have not yet solved the technical bottleneck of intelligent collaboration among large-scale groups. as a new paradigm that surpasses individual intelligence, swarm intelligence technology, as a way to stimulate large-scale participants to complete real and complex tasks, needs to be explored. This paper focuses on shared travel, which is one of the most typical and practical applications of swarm intelligence technology. Firstly, it discusses several frontier issues of swarm intelligence technology in large-scale shared travel scenarios, including organizational structure design, behavior pattern mining, incentive mechanism design and collaborative decision optimization, and introduces the relevant cutting-edge research work. Then it takes DiDi as an application case to illustrate the real application scenario of swarm intelligence technology. Finally, the prospect of its future development is made.

Artificial intelligence; swarm intelligence; shared travel

Research background

In recent years, the development of artificial intelligence technology has undergone drastic changes, but also made a major breakthrough. We have seen a series of eye-catching technological innovation products led by AlphaGo, but at the same time, we also realize that the current development of artificial intelligence is facing a bottleneck that is difficult to land. The reason is that the mainstream paradigm of artificial intelligence often puts too much emphasis on the simulation of individual intelligence, which is finally limited to solving the problems of closed boundary, static constraint and single goal. In real life, with the high-speed popularity of the Internet, billions of netizens are colliding, communicating and connecting data and information every day. In this environment, the boundary of intelligent system changes from closed to open, the constraint of intelligent construction changes from static to dynamic, and the goal of intelligent application changes from simple to complex (see figure 1). In order to depict this new intelligence paradigm, swarm intelligence technology came into being. The so-called swarm intelligence refers to the intelligence that emerges beyond individual intelligence when large-scale participants are attracted, gathered and managed to complete tasks through a specific organizational structure. The idea of swarm intelligence has existed since ancient times. As early as the ancient Greek city-state era, people began to gather in the square to discuss politics by voting by many people. Since ancient times, there is also a proverb in our country that "everyone gathers firewood and the flame is high". Since modern times, with the development of system science and electronic computer, there have been many early researches with the idea of swarm intelligence. such as the concept of group intelligence put forward by Craig Reynolds, the hall system of comprehensive integration initiated by Qian Xuesen, the multi-agent system put forward by Victor Leiser, and the crowdsourcing technology which has attracted more and more attention in recent years. Inspired by the traditional idea of swarm intelligence and cutting-edge artificial intelligence technology, a new set of Internet-based swarm intelligence technology is gradually taking shape.

As a general term of a kind of new technology, the specific research content of swarm intelligence technology can not be discussed without the support of application. Swarm intelligence technology is widely used in many fields, such as military reconnaissance, prevention and control of infectious diseases and social behavior analysis. This paper will focus on the most typical of them, but also the most related to people's daily life, that is, shared travel. The so-called shared travel refers to a new mode of transportation in which social groups share vehicles in the same or different periods of time and pay corresponding fees according to their own travel requirements. Its representative companies, such as Uber and DiDi, have various types of shared travel services, such as DiDi's express, chauffeured and ride-sharing services. With the popularity and development of mobile Internet, the scale of groups involved in shared travel begins to expand. According to statistics, by 2019, the total number of DiDi users has exceeded 550 million, transporting 10 billion passengers a year. In such a large-scale real shared travel scene, how to more effectively motivate and coordinate group participants (such as passengers and drivers) in order to provide people with more convenient and efficient travel services has become the focus of research. swarm intelligence technology is the key to solve this kind of problem. This paper first focuses on several frontier issues and research status of swarm intelligence technology for shared travel, and then takes DiDi as an application case to illustrate the real application scenario of swarm intelligence technology. finally, the existing research problems and future research challenges are discussed.

1 Frontier issues

As mentioned above, the swarm intelligence technology for shared travel mainly studies how to optimize travel services through passive or active ways such as motivating and regulating groups; at the same time, effective incentive and regulation means are also inseparable from the understanding of group organization and behavior as a support. Therefore, four frontier issues are discussed in detail here. The first is how to design and optimize the organizational structure of the group. An efficient organizational structure is a prerequisite for the emergence of group intelligence. In shared travel, a reasonable organizational structure for the driver group can often stimulate higher travel service efficiency. The second is how to mine and predict the behavior patterns of the group. Participants in shared travel, both drivers and passengers, have the characteristics of autonomous and uncertain behavior, so the establishment of reliable prediction models for group behavior can often help decision makers to better coordinate and manage large-scale groups. group behavior patterns have also become an indispensable part of the research. There is a close relationship between organization and behavior, the organizational form of the group determines the research category of the group behavior pattern, and the prediction of the group behavior also affects the design method of the group organizational structure. On the basis of designing the organizational structure and understanding the behavior pattern, the third frontier problem is how to design a desirable incentive mechanism to guide the group. As rational individuals, drivers and passengers often sacrifice the interests of the group (such as global income) in order to maximize their personal interests (such as revenue from a single order). The function of incentive mechanism is to provide rational incentives for groups to guide them to achieve the design goal of group intelligence system. It is an important link to attract and gather drivers or passengers to participate continuously in order to achieve the emergence of group intelligence. At the same time, the individuals of the group intelligence system are not completely autonomous and uncontrollable, for example, the platform can dispatch orders to drivers, so in addition to the incentive mechanism, the fourth is how to make efficient collaborative decision-making for the group. it aims to provide critical and continuous decision support for groups in the process of achieving their goals, so as to finally achieve the goal of the emergence of group intelligence. The incentive mechanism is as inseparable as collaborative decision-making, which constitutes a loop feedback relationship in which the decision is influenced by the incentive mechanism, and then the decision result is fed back to the incentive mechanism, so as to update and optimize. The following focuses on these four frontier issues in more detail.

1.1 organizational architecture design

The traditional theory of multi-agent system has long been involved in the study of organizational architecture, such as designing a hierarchical organizational architecture to achieve efficient task allocation, information transmission and result aggregation when multiple robots cooperate to complete a task. However, in the field of shared travel in the Internet environment, the agent has changed from a simple and controllable robot to a group of drivers or passengers with strong autonomy, complex characteristics and heterogeneous members. its number has also changed from no more than 10 to tens or even millions, and members flow in real time, and its dynamics is far beyond the scope of the traditional multi-agent system. Therefore, it brings great challenges to the optimal design of organizational structure. Which is a good http://www.hnzzzy.com/ for Zhengzhou abortion Hospital

At present, in the field of shared travel, one of the typical problems of organizational structure design is the problem of driver formation, that is, how to divide the driver group into several teams in order to complete the order task more efficiently. In the problem of driver team formation, the core point is to determine the structure and composition of the team in order to maximize the overall efficiency and other optimization objectives. For example, you need to decide whether to use a team leader with multiple team members or only the team members without the team leader structure; at the same time, you also need to decide the number of teams and which drivers are arranged to join which teams. In order to solve the optimization problem of organizational structure, it is necessary to establish a comprehensive evaluation system for the contribution of different members of the team, such as the order acceptance efficiency of each driver in different time periods, so as to form the relationship between the individual and the overall goal. Then, the constraint relationship of complex preferences among individuals needs to be established. for example, many drivers prefer to team up with their fellow villagers or drivers with similar interests. At the same time, we should also realize that the optimal team organization structure solved on the basis of the above work is only suitable for static scenarios, while ignoring that drivers have the characteristics of dynamic flow and uncertain online and offline time. Therefore, finally, it is necessary to design dynamic evaluation strategies and highly robust control measures for possible emergencies of key nodes in the organization. At present, DiDi has made preliminary achievements in the static scene on the problem of driver formation, but many problems have not been solved, and there is still a lot of research space for the design of group organization structure in the future.

1.2 behavior pattern mining

The research of behavior pattern mining and the design of group organizational structure complement each other. a variety of contribution and preference evaluation methods adopted by organizational structure are inseparable from the understanding of behavior patterns. at the same time, the prediction of behavior can also react to the regulation of organizational structure. In a general sense, the mining and analysis of human behavior patterns has always been one of the hot and important research directions in the field of traditional data mining and machine learning. Driven by the development of mobile Internet and Internet of things, smartphones, smart wearable devices and street cameras produce a large amount of human behavior data every day. Some existing research fields, such as mobile swarm intelligence perception and spatio-temporal crowdsourcing, involve the collection and analysis of such behavior data, and many of their methods can also be used for reference in the research of swarm intelligence. However, in the broader research of swarm intelligence technology, there are a variety of dynamic data sources, the objectives are vague and diversified, the relationships between group behavior patterns are often more deep and complex, and there are many limitations, such as privacy protection. Therefore, it also brings a certain degree of challenges to the research.

In the field of shared travel, the understanding of behavior patterns can be divided into driver behavior and passenger behavior. The driver's behavior includes the departure time period, the habitual work area and the average daily income, etc., in addition, the driver's personal portrait and passenger preferences can also be included in the scope of his behavior. The work related to the understanding and prediction of driver behavior generally uses the historical data of driver behavior to establish a machine learning model to predict the future behavior of drivers, so that the prediction results assist other processes such as team formation, order dispatch and so on. Passenger behavior can also include passengers' daily taxi-hailing habits and taxi-hailing preferences, which can be modeled and predicted by similar methods. The most typical behavior pattern mining problem in shared travel is supply and demand forecasting, which aims to establish a model based on a large number of historical order data to predict taxi supply and demand in different regions and periods of the city in the future. as a result, drivers are dispatched in advance to alleviate the imbalance between supply and demand. Part of the representative existing work is devoted to comprehensive feature engineering and the implementation of distributed algorithms for massive data. However, nowadays, many behavior patterns mining in the field of shared travel involves a large amount of personal data of drivers or passengers, which is faced with more and more privacy restrictions in the process of data collection. How to model and predict machine learning under the premise of protecting personal data privacy is not only the focus of federal learning, but also a major challenge in swarm intelligence behavior pattern mining in the future.

Thank you for your reading. the above is the content of "Analysis of Swarm Intelligence Technology and Application for shared Travel". After the study of this article, I believe you have a deeper understanding of the problem of analyzing the technology and application of swarm intelligence for shared travel, and the specific use needs to be verified in practice. Here is, the editor will push for you more related knowledge points of the article, welcome to follow!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.