Design Of Server-side Software For Immersed Echo Simulator

In the fields of transportation,geographic mapping and national defense,electromagnetic echo detectors play a very important role and shoulder the important mission of real-time monitoring.Electromagnetic echo simulator is widely used in the test and verification of detection equipment and related engineering.Electromagnetic echo simulation requires high real-time,throughput,interactivity and availability of the simulator because of its complex scene setting and short inter-frame interval.Traditional simulators often use isomorphic platforms and stand-alone implementations,and weaken the ability of human-computer interaction.According to the above requirements and some shortcomings of the current simulator,the following work has been done in this thesis:1)Simulation and acceleration of electromagnetic echo signals.The simulator's simulation algorithm is complex and computationally intensive,and it takes a lot of time in the traditional CPU isomorphic platform.This thesis builds a CPU-GPU heterogeneous platform,studies the underlying OpenCL mapping implementation of AMD GPU,and optimizes the implementation of simulation algorithms from the aspects of parallel computing,memory access,and data transmission.This thesis uses cluster to improve the throughput of the system,and use reverse proxy and task queue feedback mechanism to achieve load balancing.2)Client/server architecture design.This thesis realizes the sharing and synchronization of multiple clients to the same scene and collision detection between targets to meets the needs of large-scale drills.On the client server communication,the Reactor mode-based I/O multiplex communication component is designed to save thread resources and improve communication efficiency.3)System stability design.In this thesis,the single-point system in the system uses the master-slave structure and uses the IP drift technology to achieve automatic switching between the master and the backup.For the machines in the cluster,use the heartbeat package and sampling verification to perform health check and automatically remove the abnormal machine.At the same time,the centralized log system is built to establish a monitor to realize the visualization of cluster resources.Finally,the server-side software of distributed echo simulator supporting multiple clients is designed and implemented.The comprehensive test results show that the simulator designed in this thesis can accurately simulate real-time signals under real-time requirements,and can automatically switch between master and slave fault scenarios.Besides,it has good synchronization and interaction with the client.In terms of performance,the CPU-GPU heterogeneous computing platform designed in this thesis achieves an acceleration ratio of 73.8 compared with the CPU isomorphic computing platform.On the basis of introducing a small amount of communication loss,computing cluster can improve the throughput of signal processing of the whole system almost proportional to the number of computing cluster machines.