Vehicle state estimation method based on adaptive volume particle filtering

Abstract

The invention discloses a vehicle state estimation method based on adaptive volume particle filter (ACPF), which comprises the following steps: firstly, constructing a high-dimensional nonlinear eight-degree-of-freedom vehicle dynamic model based on an unsteady state dynamic tire model; secondly, updating an importance density function of a basic particle filtering algorithm by utilizing an adaptive volume Kalman filtering algorithm so as to complete the design of the adaptive volume particle filtering algorithm; achieving high-precision online observation of key state variables such as roll angle and side slip angle of a vehicle by using vehicle-mounted sensor information and an ACPF algorithm; finally, building a Simulink-Carsim joint simulation platform for algorithm verification. The result shows that the algorithm state estimation precision is higher than that of a traditional unscented particle filter (UPF) algorithm, and the algorithm operation efficiency is higher than that ofthe UPF algorithm.

Description

technical field [0001] The invention relates to a vehicle state estimation method based on an adaptive volumetric particle filter (ACPF), and belongs to the technical field of vehicle monitoring. Background technique [0002] With the rise and popularity of smart cars, the requirements for vehicle dynamics-based vehicle control are getting higher and higher, which inevitably involves accurate identification of the state of the vehicle itself. When the driver is driving a car, he can perceive certain state variables of the vehicle, such as the body roll angle, but the smart car controller cannot do this. It is particularly important to visualize the data of the necessary state parameters required for dynamic control. [0003] At present, the main methods of vehicle state estimation include Kalman filter, particle filter, fuzzy logic method, neural network method, synovium observer method, Luenberger observer method, robust observer method, and the classical Kalman filter (KF...

AI Technical Summary

Problems solved by technology

But there are various problems: it is difficult to determine the proportional factor and quantization factor in the fuzzy logic method, the neural network method depends on the performance of sensors and processors, there is chattering phenomenon in the state estimation of the synovial film observer method, and the Luenberger observer method is in some In some cases, there is an underestimated estimation bias. EKF needs to perform first-order Taylor expansion on the state equation, which reduces the estimation accuracy. UKF and CKF do not need to calculate the Jacobian matrix of the nonlinear function, which improves the estimation accuracy, but it may be possible in non-Gaussian system state estimation. Divergence occurs; particle filters show good performance in dealing with nonlinear non-Gaussian system state estimation, but are prone to particle degradation problems due to the difficulty in choosing the importance density function

Images