基于蚁群优化的边缘化粒子滤波器跟踪算法

An improved Marginalized Particle Filter algorithm based on ant colony optimization for tracking

针对边缘化粒子滤波器(MPF)中的Kalman滤波通常无法从量测方程中获得线性状态估计,以及标准粒子滤波器估计非线性状态计算量大和粒子退化的问题,提出了一种改进的MPF粒子滤波目标跟踪算法。该方法采用状态的预测值作为Kalman粒子滤波的量测更新,用Kalman滤波估计目标的速度和加速度,并通过蚁群优化算法改进标准粒子滤波器,并用以估计目标的位置信息。仿真结果表明:该方法在提高目标状态估计精度的同时,降低了粒子滤波的计算复杂度,克服了粒子滤波退化现象。

Aiming at the problem of marginalized particle filter（MPF） in which the Kalman filter（KF） is usually hard to obtain the linear state ,and the traditional particle filter （PF） usually exists the deficiency of high computation complexity and particle degeneracy .An improved M PF algorithm for maneuver target tracking is proposed .The new algorithm ,in one hand ,uses the forecasting values of state as the measurement updating of Kalman filter and then uses KF to estimate the speed and acceleration ,in the other hand ,the new algorithm can improve the traditional PF by ant colony optimization algorithm to estimate target position .Simulation results indicate that the improved M PF algorithm can not only improve the estimate precision of target states but also at the same time reduce the computation complexity and overcome degeneracy phenomenon of the traditional PF .