The Application Of Improved Particle Filter Algorithm In The Interactive Of Multiple Model

Classic filtering used in tracking is the Kalman Filter, which is the optimal linear estimation under Gaussian noise or filtering. But it can't be used in non-linear ane non-Gaussian noise environment, and even lead to divergence, so It is improvied to be extended Kalman filter. In order to better Tracking objects in a nonlinear environment, in recent years there has generated a lot of new filtering algorithm for nonlinear environment, including the particle filter. Classical particle filter has many disadvantages, such as easily dissipate, deprivation and other particles. When particle filtering algorithm is improved, a variety of improved particle filter algorithm generates, such as extended Kalman filter, Unscented Kalman Filter algorithm, adaptive particle filter particles, MCMC particle filter, interacting multiple model particle filter algorithm. Filter algorithm used in target tracking have so many advantages and disadvantages of each, in different environments each tracking filter algorithm has different filtering effect.In this paper, target tracking model will be described detailly which is often used in algorithms such as CA/CV model, the current statistical model and the interacting multiple model algorithm. Kalman filter, Kalman filter algorithm and particle filter will be introduced too.It is introduced in many paper that sampling number of particles, the process noise and measurement noise on the particle filter will have a huge impact in the particle filter filtering results, this paper will simulate to verify this point, and the sampling number of particles, the process noise and measurement noise filtering particle filter will effect what kind of impact. Be same the nonlinear filtering algorithm, compared to the extended Kalman filter,particle filter algorithm has great advantages, and its computational complexity is not the Dimension equation, the algorithm complexity is related only to the number of particles. However, the filtering effect of particle filter is related to the sampling number of the particles, when the number of particles sampled is low, filtering effect of particle filter may be worse than the extended Kalman filter. This paper will compare filtering effect of particle filter algorithm with extended Kalman filter when particle filter take different number of particles ofA variety of papers on the particle filter has mentioned to set a sampling threshold and it is only resampling when the effective number of particles less than this value and made it clear that this will reduce the number of re-sampling. But set the threshold will be more steps to calculate the threshold, so that it will increase the complexity of the algorithm to extend the completion time of the algorithm, and the other because the effective sample size at the threshold higher than the sample without resampling, particles can also cause a degree of degradation phenomenon, it will also affect the final filtering.In the improved particle filter algorithm, the extended Kalman particle filter, a regular particle filter, particle filter and optimal linear filtering particle filter will be compared, and the best of the particle filter algorithm is applied to improve the interactive multiple model algorithm,last the filtering effect of this algorithm will be researched.