Particle Filter Algorithm For Attitude Estimation With UAVs And Its FPGA Implementation

Mini Ummanned Aerial Vehicles(UAV)have been greatly developed in the military and civilian fields for their low cost and driverless features.For mini UAVs,accurate and real-time positioning and navigation plays a crucial role in flight safety and reliability.In the positioning and navigation,the key step of UAV navigation solution attitude estimation is difficult to obtain by the traditional nonlinear Kalman filtering method.The particle filter using probability statistics method has a great ability for dealing with non-linear non-Gaussian problem with large amounts of data and high precision requirement.Combines the advantages of Field Programmable Gate Array(FPGA)like flexibility and parallelism,this paper gives the simplified particle filter algorithm and the improved Gaussian particle filter algorithm,which can effectively apply the design to attitude estimation with UAV and also has a large of practical value for other nonlinear estimation problems.The main contents are as follows:Firstly,the commonly used navigational coordinate and the attitude description parameters are given,and the theoretical derivation of SINS attitude solving system with quaternion description is carry out.The quaternion and geodetic velocity updates under discrete conditions are shown as equation.Secondly,the methods of modeling and processing of the dynamic signals are introduced.Based on the theory and method of Bayesian estimation,the relevant principles and methods of particle filtering are given.Taking the propagation equation of the discrete quaternion and the geodetic velocity as the sampling strategy with the generalized particle filter algorithm,we can get the pseudo-code of the attitude estimation algorithm,and deduce the "average" method of quaternion.Thirdly,in order to implement the attitude estimation algorithm on FPGA,the resampling algorithm and the weight calculation step can be simplified in case the performance of the simplified particle filter algorithm is not obviously reduced.The simplified attitude estimation algorithm is decomposed into four modules as follows: Sampling Particles,Weight Calculation,Resampling and State Estimation.Using control module complete the timing control and synthesize these modules,then UAV attitude estimation algorithm based on SYSTEM GENERATOR is carry out.The results of logic simulation and numerical simulation verify the feasibility and effectiveness of this design.Finally,the simplified resampling algorithm of particle filter is still complex and not easy to operat parallelly.For Gaussian particle filter need not resampling,therefore it is used to attitude estimation algorithm design.In order to achieve pipeline operation,non-standardized weights are used to calculate the quaternion "average" value and covariance matrix,a new covariance matrix calculation formula is also carry out to improve the Gaussian particle filter.Simulation results show that the improved algorithm is feasible and effective,and the remaining resources of FPGA can be used to further improve the running speed.