一种自适应免疫优化的无迹粒子滤波器

Unscented particle filter algorithm based on adaptive immune optimization

针对无迹粒子滤波(UPF)在较偏观测时的退化现象及重采样带来的粒子枯竭问题,提出一种自适应免疫优化的无迹粒子滤波算法(AIO-UPF)。该算法在重采样过程中,利用免疫算法在亲和度与浓度调节机制下的全局寻优能力和多样性特征,通过引入自适应阈值因子δ的Metropolis准则,使得粒子集能够有效地分布于高似然区域,提高了粒子的多样性和有效性,从而较好地抑制了在较偏观测时的粒子退化问题。仿真结果表明,AIO-UPF的性能优于传统UPF及标准粒子滤波,在状态估计精度上比传统UPF提高了27%左右。

Aiming at the problem of Unscented Particle Filter（UPF） such as particles degeneracy and particles impoverishment at the partial observation, this paper proposes an Adaptive Immune Optimization Unscented Particle Filter （AIO-UPF） algo- rithm. The algorithm uses the global optimization ability and diversity of features of the immune algorithm in the affinity and concentration and the Metropolis criteria with the adaptive threshold factor δ makes the particle set move towards higher likeli- hood area. In this way, the diversity and effectiveness of particte have improved and the problem of particle degradation and de- pletion have alleviated. Simulation results indicate that the new particle filter outperforms obviously superior to PF and tradition- al Unscented Particle Filter, and in the state estimation accuracy increases by about 27%.