Study Of Improved Particle Filter Algorithm And Its Application

In real society,precise position information can be achieved by satellite navigation system.Radar and infrared system can be used to achieve target tracking where satellite signal is lost or cannot work effectively.Target tracking plays an important role in many fields,such as submarines,missiles in military field and automobile driving in civil field.The application scenes is mainly non-linear and filled with various noises in real society.The aim of estimation is to recover the original signal state from the observation signal state mixed much noise.In summary,the relative research of estimation plays a profound role in the real society.Particle filter algorithm based on Monte Carlo method has been used in non-linear estimation widely,so it has become one of the research hotspots in non-linear estimation field.The thought of resampling is introduced to particle filter to solve particle degradation,which improves the precision of filtering.However,the mechanism of resampling which uses high weight particles to replaces low weight particles caused lack of particle types in essence.The loss of particle diversity is caused by resampling algorithm,An important research point is to increase diversity of particles.First of all,fundamental resampling method,including polynomial resampling(random resampling),systematical resampling,layered resampling and residual resampling are discussed in the process.Matlab software is mainly used to achieve the above resampling algorithm.Then the effects of resampling is discussed by numerical experiments.At last,one of resampling methods is used to the next research.The problem of lack of particle diversity is inevitable by using resampling algorithm.Swarm intelligence algorithm is deeply studied and firefly algorithm is selected to modify and apply.An improved particle algorithm based on optimal neighbor to guide firefly movement is proposed.Firstly,the latest observation value is introduced into the calculation of the relative brightness of fireflies.Secondly,the decrease of distance caused the increase of relative attraction,which makes individual oscillation near the optimal individual.A decreasing function is introduced to update the attractiveness equation.Thirdly,the optimal neighbor is used to guide the individual firefly to move and control the search range.Finally,performance of the proposed algorithm and particle filter algorithm under different noise conditions is compared through three tracking experiments of distance,angle,distance and angle.The simulation results show that the tracking effect of the improved algorithm is better than that of the particle filter algorithm.