The problem of high energy consumption for mobile devices in mobile edge computing is becoming increasingly prominent. In order to reduce the energy consumption of the mobile devices, an Energy-aware computation Offloading for Workflows (EOW) was proposed. Technically, the average waiting time of computing tasks in edge devices was analyzed based on queuing theory, and the time consumption and energy consumption models for mobile devices were established. Then a corresponding computation offloading method, by leveraging NSGA-Ⅲ (Non-dominated Sorting Genetic Algorithm Ⅲ) was designed to offload the computing tasks reasonably. Part computing tasks were processed by the mobile devices, or offloaded to the edge computing platform and the remote cloud, achieving the goal of energy-saving for all the mobile devices. Finally, comparison experiments were conducted on the CloudSim platform. The experimental results show that EOW can effectively reduce the energy consumption of all the mobile devices and satisfy the deadline of all the workflows.

As the limitation of single feature representation may cause low accuracy of fine-grained image classification, a multi-feature combination representation method based on Convolutional Neural Network (CNN) and Scale Invariant Feature Transform (SIFT) was proposed. The features were extracted from the entire target, the key parts and the key points comprehensively. Firstly, two CNN models were trained with the target-entirety regions and the head-only regions in the fine-grained image library respectively, which were used to extract the target-entirety and the head-only CNN features. Secondly, the SIFT key points were extracted from all the target-entirety regions in the image library, and the codebook was generated through the K-means clustering. Then, the SIFT descriptors of each target-entirety region were encoded into a feature vector by using the Vector of Locally Aggregated Descriptors (VLAD) along with the codebook. Finally, Support Vector Machine (SVM) was used to classify the fine-grained images by using the combination of multiple features. The method was evaluated in CUB-200-2011 database and compared with the single feature representation method. The experimental results show that the proposed method can improve the classification accuracy by 13.31% compared with the single CNN feature representation, which proves the positive effect of multi-feature combination on fine-grained image classification.

Aiming at the problem of unbalanced utilization rate distribution of charging station caused by disordered charging for a large number of electric vehicles, an orderly charging strategy for electric vehicles was proposed. Firstly, the location of the electric vehicle's charging demand was clustered, and the hierarchical clustering and quadratic division based on K-means were used to achieve the convergence of electric vehicles with similar properties. Furthermore, the optimized path to charging station was determined by Dijkstra algorithm, and by using the even distribution and the shortest charging distance of electric vehicles as objectives functions, the charging scheduling model based on electric vehicle clustering was constructed, and the genetic algorithm was used to solve the problem. The simulation results show that compared with the charging scheduling strategy without clustering of electric vehicles, the computation time of the proposed method can be reduced by more than a half for large scale vehicles, and it has higher practicability.

To address the problems of large energy consumption and illegal possession of computing resources by users in cloud environment, a new algorithm named Fair and Green Resource Scheduling Algorithm (FGRSA) was proposed to save resources and enhance the fairness of the system, so that all users can reasonably use all the resources in the data center. By using the proposed method, various types of resources can be can scheduled to make use of all resources to achieve relative fairness. The simulation experiments of the proposed scheduling strategy was conducted on CloudSim. Experimental resutls show that, compared with Greedy algorithm and Round Robin algorithm, FGRSA can significantly reduce the energy consumption and simultaneously ensure fair use of all types of resources.

Concerning the problems of resource allocation about uneven distribution, low efficiency, dislocation and so on, a new algorithm named Global Dominant Resource Fair (GDRF) allocation algorithm which adopts several rounds of allocation was proposed to meet the needs of different users, achieve multiple types of resource fairness, and get high resource utilization. First, a qualification queue was determined by allocated resource amount of the users, then the specific user was determined to allocate resource through the global dominant resource share and the global dominant resource weight. The matching condition of resources was took into account in allocation process and the progressive filling of Max-Min strategy was used. In addition, the universal fairness evaluation model of multi-resource allocation was applied to the specific algorithm. Comparison experiments were conducted based on a Google's cluster. Experimental results show that compared with maximizing multi-resource fairness based on dominant resource, the amount of allocated virtual machine is increased by 12%, the resource utilization is increased by 0.5 percentage points, and fairness evaluation value is increased by about 15%. The proposed algorithm has a high degree of adaptation of resources combination allocation, allowing the supply to better match users' demand.