Research On Fast Solution Methods Of Ship Reachable Set

In recent years,with the increasing attention of the country to the shipping field,the safety issue of sea travel has become increasingly important.Ship reachable set can be used in the research of ship collision avoidance and risk assessment.Ship reachable set refers to the set of all terminal states that can be reached by the system under certain control conditions given the initial state of the ship.The calculation of ship reachable set has a large amount of data,a long time and a lack of real-time performance.High-performance parallel computing of ship reachable set can effectively avoid the above problems and realize the effect of fast calculation with real-time performance.At the same time,by learning the ship reachable set model and fitting the ship reachable set under other control input conditions using the neural network fitting method,the time-consuming problem caused by the iterative solution of numerical differential equation in the traditional ship reachable set solution method is optimized.Therefore,this thesis adopts parallel computing based on GPU,CPU-GPU heterogeneous computing,neural network approximate fitting calculation,and neural network fitting in CUDA environment to calculate ship reachable set.The main work of this thesis is as follows:Firstly,the research background and significance of reachable sets,parallel computing and fast solving methods of differential equations are described,and several numerical solving methods of reachable sets are reviewed.Based on the differential equation given by the ship dynamics model,Euler method is used to solve the differential equation,and the ship trajectory is obtained and the ship reachable set is generated.The ship reachable set is solved in MATLAB 2018 a environment and Visual Studio 2019 environment.In order to verify the accuracy of the solution results,the difference between the results under the two calculation modes is compared,and the solution error is within the allowable range under the C++development environment.Secondly,to solve the problem of long running time due to the large amount of solving data for the traditional ship reachable set,the parallel computing method in CUDA environment is adopted to solve the ship reachable set.Simulation experiments show that the computing time of ship reachable set in CUDA environment is greatly reduced compared with that in MATLAB environment,and the calculation rate is over 240 times.Since CPU threads are idle when CUDA is running,resulting in a waste of thread resources,a dynamic scheduling thread algorithm based on CPU-GPU heterogeneous mode is proposed to calculatethe ship reachable set,and the running time of ship reachable set in the heterogeneous environment and CUDA environment is compared.It is found that the calculation time of ship reachable set in heterogeneous mode is increased by 1.2 times.Finally,this thesis proposes a method for approximate fitting of ship’s reachable set based on neural network,optimizes the time-consuming problem caused by iterative solving of ship’s reachable set based on time variable and rudder Angle variable using numerical differential method in MATLAB environment,and grafts neural network to CUDA environment for approximate solving of ship’s reachable set.Compared with CUDA environment,the parallel computing time is accelerated by 1.4 times.In summary,in the research work of the fast solution method based on ship reachable set,the fast and accurate calculation of ship reachable set has laid a good foundation for the actual rapid calculation of the state of ship field in various situations in navigation.