| The location of the mobile robot is now a hot research, the Kalman filter is a commonly used method, but it usually requires a system of linear Gaussian model, so it is difficult to the needs of the actual work. In this case, the particle filter is much attention because of its adaptability.However, in practical application, there are also some drawbacks of the particle filter. For example, the introduction of resampling is to solve the degradation problem, but it also brought a problem of diversity loss. As we can know from filter theory, the accuracy of the algorithm increases along with the number of particles increases, but it also increases the computational complexity and affects the real-time performance of the algorithm.In this paper, we improve the particle filter. First, we use a stratification method and a linear combination of particles to generate new particles in resampling process, it solve the problem of particle diversity loss. Then, using the JS distance to measure the similarity of particles sets, so that the number of particle changes adaptively. When the number of the particles is too big, it reduces the computational complexity of the algorithm, and we verify the effectiveness of this method through experiments. Finally, the improved algorithm is applied to the localization of mobile robot, and get a good accuracy. |
