強健補償演算法於輪胎力估測器及電動車之整合應用

Abstract

本研究將一種強健補償演算法應用於輪胎力估測器中，並將其估測結果應用於車身穩定系統中。其演算法內容為整合遞迴型神經網路的卡爾曼濾波器，其目的在於增加估測器的強健性及降低對參數的靈敏度，以及在實車試驗時減少雜訊的干擾，同時增加車身穩定控制應用之強健性。 在輪胎縱向力估測方面，以輪胎剛體的單輪受力圖作為基礎，使用滑模估測器來做輪胎縱向力的估算。此估測器在行車過程中將提供估測值給整合車身穩定控制做判定依據；在側向力估測方面，以Magic Formula的輪胎模型，並以Hybrid Levenberg–Marquardt method and quasi Newton(LMQN)的一套非線性最小二乘方演算法來做輪胎側向力的估算；而在正向力估測方面，使用了車體動態模型進行方程式推導。其中，車身穩定控制包含直接偏擺力矩控制器(direct yaw-moment controller, DYC)及防鎖死煞車系統(anti-lock braking system, ABS)。DYC藉由側滑角速度及偏擺角速度，以β-γ相位穩定圖判斷車輛穩定性，再整合滑模控制及PSO粒子群最佳化法即時分配各馬達之驅動力矩，使車輛轉向時依然能保持車輛的轉向穩定性；而ABS能根據駕駛者的煞車命令即時分配各馬達之煞車力矩，且利用積分型滑模控制調整車輛之煞車油壓以防止輪胎打滑與失控。 本研究以模型迴路(model-in-the-loop, MIL)及實車試驗驗證輪胎力估測器及強韌補償器之性能。實驗以本實驗室之多動力馬達電動車作為模型架構，實車採用15-kw直流無刷馬達搭配傳動齒輪箱，作為前輪之間接驅動動力源；後輪則由兩顆7-kw永磁同步馬達至於輪內，作為後輪之直接驅動動力源。此架構能藉由操作各馬達的輸出力矩於高效率區間，達到提升整體行車效率與續航力之效果。

In this research, a robust compensation algorithm is applied to the tire force estimator, and its estimation results are applied to the vehicle stability system. The content of the algorithm is a Kalman filter integrating a recursive neural network. Its purpose is to increase the robustness of the estimator and reduce the sensitivity to parameters, as well as to reduce the interference of noise during the actual vehicle test, while increasing the robustness of the vehicle stablility control applications. In the aspect of tire longitudinal force estimation, the estimation of longitudinal tire force is based on the single-wheel force diagram of the rigid body of the tire, and the sliding mode estimator is used. This estimator will provide the estimated value during the driving process for the judgment of the integrated body stability control; in terms of lateral force estimation, it uses the tire model of Magic Formula and uses the Hybrid Levenberg–Marquardt method and quasi Newton (LMQN), a set of nonlinear least squares algorithm, as the algorithm of the estimation of the tire lateral force; and for the normal force estimation, the dynamic model of the vehicle body is used to derive the equation. Among them, the vehicle stability control includes direct yaw-moment controller (DYC) and anti-lock braking system (ABS). DYC judges the stability of the vehicle by using the β-γ phase stability diagram based on the side slip angular velocity and the yaw angular velocity, and then integrates sliding mode control and PSO particle swarm optimization method to instantly distribute the driving torque of each motor, so that the vehicle can still be turned Maintain the steering stability of the vehicle; and ABS can instantly distribute the braking torque of each motor according to the driver's braking command, and use integral sliding mode control to adjust the vehicle's braking oil pressure to prevent tire slip and out of control. In this study, model-in-the-loop (MIL) and actual vehicle on-road testing were used to verify the performance of the tire force estimator and robust compensator. This composite estimator system is based on the powertrain of EV consists of three motors: a 15-kW front traction motor with gearbox and two 7-kW in-wheel motors installed inside both rear wheels. This configuration not only provides the vehicle good performance for planar motion control, but also improves driving efficiency of vehicle by driving and regenerative braking torque distribution of three motors.