This project belongs to the field of computer science. Mobile network optimization is a research frontier and hot topic in related fields internationally. However, topology dynamics, time-varying resources, service complexity and user diversity make time-varying resources difficult to represent, heterogeneous networks difficult to coordinate, and complex networks difficult to optimize, making mobile network optimization a challenging problem. Solving this problem will provide theoretical support for my country to achieve a network power. From 2006 to 2016, with the support of the National Natural Science Foundation of China's Major Research Program, the 863 Program, and Pujiang Talents of Shanghai City, this project is based on multi-dimensional cognition, with adaptive collaboration as the core, and with mobile network optimization as the goal. Systematic original research has been carried out and a set of mobile network optimization theories and methods have been formed. The main scientific findings are as follows: 1. The cognitive mechanism of time-varying joint control of the network environment was discovered. The time-varying nature of resources and topological dynamics pose dual challenges to resource availability. This project establishes a dynamic control method based on resource state prediction and dynamic graph theory, proposes a time-varying joint control model of resources, and reveals the spatio-temporal evolution mechanism of mobile network resources. Sherman Shen(IEEE Fellow, academician of the Canadian Academy of Engineering), a professor at the University of Waterloo in Canada, believes that Representative Paper 6 is a typical top-k-based method for obtaining mobile network resource status. 2. An intelligent sensing method for user behavior and business characteristics is proposed. Using user portable devices, photoacoustic touch sensors, etc., a fine-grained user behavior analysis method based on neural networks is proposed. Based on multi-granularity classification, a low-overhead, high-precision service feature recognition method is proposed. Chen Chun, academician of China Academy of Engineering, believes that our paper 4 provides a representative method for modeling perceptual range (INFOCOM 2016). Several IEEE Fellow (such as P. Mohapatra, B. V. K. V. Kumar, F.Bai, M.Guizani) commented on the user behavior cognition method proposed in paper 7 in top journals IEEE TON and TMC, which also has high perception accuracy when data is missing. 3. Establish a network optimization method based on adaptive collaboration. A cooperation gain model, cooperation mechanism and joint design method within nodes, between nodes, and between networks are proposed, and a cooperative transmission model and algorithm based on adaptive cooperation are established. Aiming at the goals of network transmission capabilities, reliability, and user needs, a set of network optimization models are proposed. Professor Andrew Drozd (IEEE Fellow)'s Evaluation Paper 3: Jointly optimizes system resources from multiple dimensions such as rate adjustment, channel allocation and routing. 4. Revealing the multi-network collaboration mechanism of transmission and calculation linkage. Based on the idea of software definition and with the goal of deep integration of heterogeneous networks, a multi-controller deployment model and collaborative management and control mechanism are proposed, and a dynamic transfer model based on data dependence and resource status is proposed. Professor Cao Jiannong of the Hong Kong Polytechnic (IEEE Fellow) and others have continued to track and spoke highly of Paper 5 in IEEE TMC and INFOCOM papers many times, believing that we have proposed a powerful and universal joint analysis method of data dependence and resource state. All results were completed independently by this project team. The overall project has reached the international advanced level. Eight representative papers were cited 115 times by SCI. The completion person published 68 related papers in top international journals and conferences such as IEEE TMC, IEEE TPDS, INFOCOM, etc. He cited 2047 times and won 4 best papers at international conferences. It has received positive comments and a large number of citations from domestic and foreign colleagues, including ACM and IEEE Fellow, providing theoretical and methodological support for my country's Internet + strategy.