Response-based methods to measure road surface irregularity: a state-of-the-art review

European Transport Research Review

An Open Access Journal

Table 2 Summary of data-driven methods for road profile reconstruction function

System name/by	Machine learnings	Additional	Suspension			Vehicle model			Main parameter
System name/by	Machine learnings	Additional	P	SA	A	Q	H	F	Main parameter
[33,34,35]	ANN (NARX)	PD	✔				✔		sprung, axle, body
[36]	ANN		✔					✔	wheels and chassis
[8]	ANN + wavelet DWT)	RE(IRI)	✔			✔			sprung mass
[37]	ANN + ADV		✔	✔				✔	unsprung mass
[38]	ANN + image processing + PCA	Terrain	✔			✔			wheel acc, speed
[39,40,41]	SVM+ PCA, FWT, FFT	Terrain	✔			✔			wheel acc, speed
DNNs classifier [42]	Deep NNs			✔				✔	sprung, unsprung, rattle space
PNN classifier [43]	PNN + WPT			✔		✔			sprung, unsprung, rattle space
ANFIS classifier [44]	ANFIS			✔		✔			sprung mass
[45]	ANFIS, RLS, GMDH			✔		✔			sprung, unsprung, rattle space
ANFIS+AKF [21]	ANFIS + Kalman filter				✔	✔			sprung mass
AKF-ASTO [22]	PNN + Kalman filter					✔			sprung, unsprung
[46]	ANFIS + MOOP + NSGA-II			✔		✔			sprung mass
[2]	RF + WPT				✔	✔	✔		sprung, unsprung, speed
SIRCS [47]	RF + TF, decision procedure		✔			✔			unsprung mass
[48]	Independent Component Analysis		✔			✔	✔	✔	chassis, suspension
[49]	Various MLs + TF	PD	✔					✔	axle or body, speed

PCA, WPD, WPT, DWT, FWT: Principal Component Analysis, Wavelet Package Decomposition, Wavelet Package Transformation, Discrete Wavelet Transform, Fast Wavelet Transform.
RLS, GMDH, ADV: Recursive Least Square, Group Method of Data Handling, the mean square of unsprung mass acceleration divided by vehicle speed.