I. Intelligent Telematics and Application Research

    With the rapid development and cross-fertilization of modern information and communication technology, automobile manufacturing industry and intelligent transportation system, the Internet of Vehicles has become one of the most promising fields in the Internet of Things system with the most industrial potential and clear market demand, representing an important direction of the deep integration of informationization and industrialization. As a cross research and application field of IOT, big data and artificial intelligence, Telematics is of great significance to improve traffic safety and efficiency, build green and smart cities, and promote the integration and innovation of information industry and traffic industry. In particular, in order to promote the evolution of Telematics in the direction of intelligence and networking, and to facilitate the development of new industries such as new energy vehicles, intelligent driving and driverless, this project focuses on how to achieve heterogeneous network convergence and effectively support large-scale data services in a highly heterogeneous, highly dynamic and highly distributed Telematics environment.

    It is easy to foresee that the future development of the automotive field can be compared to the development trajectory of smart phones in the past 10 years: 10 years ago, the call and short message attributes of cell phones accounted for more than 80% of cell phone functions, but nowadays, as an intelligent mobile device, the Internet-based information collection and transmission attributes have far exceeded the call attributes. At present, with the rapid development of electric cars, intelligent assisted driving, and driverless cars, the automotive sector is rapidly rigorous from the traditional manufacturing industry to the Internet industry. It is expected that in the course of development in the next 10 years, the basic attributes of the car as a means of transportation to travel will be significantly reduced and replaced by a powerful mobile information collection and transmission node in the world of the Internet of Things. Therefore, the development of the Internet of vehicles is the core driving field of future intelligent transportation and smart cities.

II. indoor positioning technology research for location-based services

    Indoor positioning is the basic technology to realize location-based services (LBS), and it is also one of the core research directions in the application fields of Internet of Things and pervasive computing. This project focuses on the framework and algorithms of indoor positioning, and explores and researches different application requirements by integrating navigation, crowdsourcing, data mining and other intersection fields. The main research contents of the project include

    Indoor positioning system: research the characteristics and advantages of different positioning technologies (based on fingerprint, distance, vision, etc.), study the corresponding positioning system framework and positioning algorithms for application requirements (positioning accuracy, system scale, etc.), and implement the system prototype.

    Indoor positioning navigation and application: for applications such as navigation in large complex indoor environments such as shopping malls and convention centers, parking space reservation and navigation, special crowd monitoring and trajectory tracking, etc., we conduct comprehensive research on indoor positioning, sensor data fusion, map matching, crowdsourcing and other technologies to achieve different granularity of indoor navigation and trajectory tracking functions.

    Indoor trajectory mining: based on user trajectory data, combined with artificial intelligence, big data mining and other technology research, to achieve different application requirements for user portraits, recommendation systems and human flow distribution monitoring and analysis and other functions

III. Reinforcement learning based multi-intelligence automatic mine avoidance path finding research

    Reinforcement learning is an important branch of machine learning, a kind of strategy learning in artificial intelligence, also known as re-excitation learning and evaluation learning, which is developed from theories such as animal learning and parameter perturbation adaptive control. The so-called reinforcement learning refers to the learning of mapping from environmental states to actions in order to maximize the cumulative reward value that the action receives from the environment. The method differs from supervised learning techniques in that positive and negative examples are used to inform which behavior to adopt, but trial and error (trial and error) is used to discover the optimal behavior strategy. Commonly used reinforcement learning algorithms include TD (Temporal Difference) algorithm, Q-learning algorithm, Sarsa algorithm, etc. There are many real-world examples of reinforcement learning. For example, the most famous Alpha go, where a machine beats a human master in Go for the first time, and Atari, where a computer learns to play the classic game by itself, are all examples of computers updating their own behavioral guidelines through continuous attempts to learn how to play Go well, and how to manipulate the game to get high scores step by step. In addition, reinforcement learning algorithms are one of the core technologies in some of the more cutting-edge and popular applications, such as unmanned car driving.

    In this project, we will study the reinforcement learning algorithms applicable to multi-intelligent systems, and consider the design of mutual reinforcement learning algorithms among intelligences in addition to the individual reinforcement learning algorithms of intelligences themselves.

IV. Research on the optimal delivery route of swarm intelligence logistics based on trajectory data mining

    According to the data recently released by Alibaba, online Taobao retailers generate about forty million parcels per day. To meet such a huge demand for parcel delivery, traditional logistics mainly relies on methods such as investing in more delivery vehicles and increasing the frequency of delivery, but it still seems insignificant. More seriously, this logistics mode not only greatly increases the traffic pressure of the city, but also poses a serious threat to the environment on which the residents depend. How to deliver a large number of packages to customers faster and better? Amazon and other international e-commerce giants have been committed to exploring new urban parcel delivery models, such as the use of drones to deliver parcels. However, these trial logistics models have defects such as high human and material costs, making it difficult to popularize their use on a large scale.

    In order to better solve the above problems, this project proposes a new logistics model - Group Wise Logistics, i.e.: using the surplus capacity generated by cabs in response to passenger delivery tasks to assist parcel transportation, innovatively opening up a new direction for the logistics field. The project aims to utilize and optimize the resources of group intelligence to achieve the goals of reducing logistics costs, improving logistics efficiency and environmental sustainability, and ultimately generating value-added logistics benefits. The research content of this project includes: node optimization of group wise logistics network for cost minimization, edge circulation time modeling of group wise logistics network based on distributed parameter learning, and optimal group wise logistics route discovery under random request.