In recent years, with the development of Internet technology, online education has emerged. Furthermore, the global spread of COVID-19 makes it an inevitable form of education. This paper proposes an online simulation experiment for mobile communication course, based on the concept of grading. To serve for this experiment, we designed a platform with unique source information, personally configurable channel and code clone detection function. To guide interest of students during the experiment, we distribute personalized source to them and allow students to choose differentiated channel configurations. The simulation experiment results demonstrate that personalized channel configurations are effective in helping students understand the impact of channel environments on wireless communication. Furthermore, it also stirs up students' enthusiasm for learning. After receiving the code uploaded by the student on the platform, the code cloning detection carried out can distinguish the similarity of student code, thus promoting students' independent programming ability.

Radio map modeling based on the artificial neural network (ANN) has great prospects for applications in base station siting and network coverage optimization. However, the data to be processed contains a large amount of private user data, such as position and route. Moreover, it is difficult for field test users to have detailed information about the transmission environment, such as building heights and contours. In order to solve the above problems, this paper proposes a Distributed learning-based Radio map Modeling method with environment Feature Augmentation based on the Augmented Scatter graph of field test data (DRM-FAAS). In the proposed method, each local user augments the field test data with k-Nearest-Neighbor (KNN) and obtains 15-D augmented environmental features based on target detection of scatter graphs as network input to collaboratively predict the received signal strength (RSS) of all positions in the modeling region. The results on the test dataset show that the modeling accuracy of the proposed method is improved by 13.81 and 4.34 dB, respectively, over the traditional propagation model and the method without environmental feature augmentation. Besides, comparable accuracy is achieved with the modeling method with detailed environmental information.