Particle filtering is an important tool for information fusion and has received wide attention in various fields such as

Particle filtering is a crucial tool for information fusion and has garnered significant attention in various fields, such as the Internet of Things and intelligent traffic management systems [1]-[3]. Acquiring vehicle localization information is fundamental to monitoring vehicle operation status and workshop cooperative control. However, the core difficulty lies in the uncertainty of target motion [4]. Particle filtering is well-suited to address the nonlinear and strongly coupled characteristics of the target [5]. In 1993, N.J. Gordon [6] and others proposed a new Bayesian state estimation algorithm that is nonlinear/non-Gaussian and not limited by linear or Gaussian noise assumptions. This algorithm can be applied to any state transition or measurement model, laying the foundation for the classical particle filter (PF) [7]-[10]. The PF compensates for the inability of the Kalman filter to predict nonlinear systems by directly using particle samples to approximate the probability distribution. However, weight degradation occurs during particle evolution, leading to unsatisfactory predictions. The Kalman particle filtering algorithm (KF-PF) [11] can reduce the influence of noise on the filtering results by Kalman filtering while achieving the prediction of nonlinear and non-Gaussian systems. However, KF-PF requires some a priori knowledge of the dynamic properties of the system and the observation model; otherwise, it may lead to inaccurate estimation results [12]-[14]. To solve the problem of difficult resolution of the prior model, the extended particle filter (EPF) [15]-[18] uses the idea of the extended Kalman filter to linearize the nonlinear state transfer function and the observation model at each time step and then uses a linear Kalman filter to process it. This approach improves the estimation accuracy and stability, but the computational effort is large, and it is more sensitive to the selection of the initial state. A large number of particles are required to improve the prediction accuracy. Ling-Feng Shi [19] et al. proposed the integrated particle filtering IPF algorithm, which solves the problem that the PF algorithm relies heavily on the number of particles, reduces the randomness of the PF-related algorithm, and improves the computational speed. However, the IPF algorithm is based on the assumption that all noise obeys a Gaussian distribution, and the noise generated by vehicle motion is a nonlinear non-Gaussian system. Therefore, to satisfy the prediction accuracy while minimizing computation and improving prediction timeliness, this paper performs segmentation fitting of the multistate motion system model to realize the mathematical analysis