











【摘要】在利用隨機森林算法(RF)進行路面附著系數估計時,存在模型構建過程中特征選擇不夠優化以及決策樹集成的多樣性不足等問題。為此,提出一種基于粒子群優化算法(PSO)對RF進行改進的方法,并給出算法流程。建立路面附著系數估計RF模型,使用PSO算法用于優化RF的參數配置,包括每棵樹的特征數量、樹的數量等關鍵因素,以增強模型的多樣性和泛化能力。最后,在MATLAB/Simulink平臺上搭建了聯合仿真模型進行試驗,對比試驗結果表明:基于PSO-RF的隨機森林路面附著系數估計方法能夠克服傳統RF方法中存在的局限性,其估計精度和穩定性均得到顯著提升。
關鍵詞:路面附著系數;隨機森林;粒子群優化;狀態估計
中圖分類號:U461.5+1;TP391.9" "文獻標志碼:A" "DOI: 10.19822/j.cnki.1671-6329.20240283
Estimation of Road Adhesion Coefficient Based on PSO-RF
Huang Xun, Zha Yunfei
(Fujian University of Technology, Fuzhou 350118)
【Abstract】 When using the Random Forest (RF) algorithm to estimate the road adhesion coefficient, there are issues such as insufficient optimization of feature selection during model construction and insufficient diversity in the ensemble of decision trees. To address this issue, a method based on Particle Swarm Optimization (PSO) algorithm to improve RF is proposed, and the algorithmic process is presented. An RF model for estimating the road adhesion coefficient is established, and the PSO algorithm is used to optimize the parameter configuration of RF, including key factors such as the number of features of each tree and the number of trees, so as to enhance the diversity and generalization capabilities of the model. At last, a joint simulation model is built on the MATLAB/Simulink platform for experiments. The comparative experimental results show that the random forest road adhesion coefficient estimation method based on PSO-RF can overcome the limitations of the traditional RF methods, and both the estimation accuracy and stability have been significantly improved.
Key words: Road adhesion coefficient, Random Forest (RF), Particle swarm optimization (PSO), State estimation
0 引言
路面附著系數作為車輛主動控制系統的必要輸入參數,直接影響汽車的附著力水平,并且能間接反映車輛發生滑移的可能性,較高的路面附著系數意味著車輛滑移風險更低。因此,在汽車主動安全系統中,路面附著系數是決策控制時需要考量的關鍵因素[1]。目前,對于路面附著系數的估算技術主要分為2類:基于原因(Cause-based)的方法和基于效果(Effect-based)的方法[2]。基于原因(Cause-based)的識別方法著眼于分析影響路面附著系數的各種物理因素[3],這類方法通常采用光學或激光傳感器來監測相關變量,并構建數學模型以表達這些因素與路面附著系數之間的關系。通過測量這些關鍵參數,并應用所建立的數學模型,可以計算出路面附著系數的具體數值。……