CUDA Parallel Computing And Its Application In Aircraft Guidance Control

Modern aircraft design is a complex multidisciplinary design optimization process,which exists time-consuming and complex simulation analysis model.In the process of optimization of the aircraft trajectory,a huge amount of calculation has become a problem and bottleneck.The efficiency of the optimization algorithm determines the efficiency of the whole optimization design.As a excellent global optimization algorithm,particle swarm optimization is widely used in aircraft trajectory optimization and various parameter tuning because of its easy implementation and computational power.However,the particle swarm optimization algorithm needs to search the whole area,it is possible to make the computation time become slower in the process of aircraft trajectory optimization.Therefore,it is necessary to improve the calculation speed of the particle swarm optimization algorithm in the case of large computation.At present,the improvement of the particle swarm algorithm itself can't improve the particle swarm optimization speed too much.In recent years,graphics Processing Unit(GPU)with its highly parallel computing power and good floating-point operations and other characteristics,is widely used in general computing.In 2007,NVIDIA introduced the Compute Unified Device Architecture(CUDA),which is widely used in medical imaging,computational fluid dynamics,and environmental sciences.Therefore,this paper analyzes the feasibility of parallelization of PSO algorithm and realizes PSO parallel computation based on CUDA.In this paper,the PID controller of each channel is designed after establishing the linear model of the four-axis aircraft.The PID parameters in the control process of the four-axis aircraft are set by the parallel PSO,and the control parameters are obtained and the optimization time is shortened.In addition,this paper studies the direct targeting method,uses this method to discretize the lunar soft landing trajectory,uses the fourth-order Runge-Kutta numerical integration method to integrate the differential equation numerically,and realizes the constraint of the performance index and the constraint of the parameterization.The constraint processing based on the penalty function method is introduced,and the lunar soft landing trajectory optimization problem is transformed into the parameter optimization problem.And use the parallel PSO algorithm to optimize its parameters,spend less time to calculate a better solution.Through the above example,the CUDA-based PSO intelligent optimization algorithm is applied,compared with the traditional CPU serial calculation,the computing time is shortened and the operation efficiency is improved.Therefore,CUDA parallel computing has a wide application potential and application space in aircraft trajectory optimization and various parameter tuning.