Table 6.

Comprehensive comparison among different TLMs.

TLM	Type	Data source	Description	Evolution param	Advantages	Disadvantages
O-1	Observation	WIM data	Actual collected data	–	Simple and efficient to apply	Can not consider the microscopic interactions between vehicles.High requirement of long-term relibility of WIM system.Post-evaluation of fatigue damage, not capable of prediction
O-1	Observation	WIM-Vision data fusion	Actual collected data	–	Precisely consistent to actual traffic loads	High requirement of long-term reliability of both WIM system and Machine Vision system (Fig. 5)Post-evaluation of fatigue damage, not capable of prediction
S-1	Simulation	WIM data	IDM simulator (with empirical param)	$v_0$, T, a, b, $s_0$	Only short-term WIM data is required	Not considerate with longitudinal and transverse motion features of actual traffic, causing large errors (Fig. 11, Table 5, 22.86% error in average)
S-2	Simulation	WIM data	IDM simulator (with calibrated param)	$v_0$, T, a, b, $s_0$	Only short-term WIM data is required. Capable of predictive fatigue damage evaluation with given time.Statistically aligned with traffic motion features at longitudinal direction.	Not considerate with transverse motion features of actual traffic, causing fatigue damage errors (Fig. 11, Table 5, 8.66% error in average)
S-3	Simulation	WIM-Vision data fusion	Proposed 2D-IDM simulator (with calibrated param)	$v_0$, T, a, b, $s_0$, ${\mu }_t$, ${\sigma }_t$	Only short-term WIM and Vision data is required.Capable of predictive fatigue damage evaluation with given time.Phenomenally (verified in Figs. 9, 10) and statistically aligned with traffic motion features at both longitudinal and transverse directions, causing fewer fatigue evaluation errors (Fig. 11, Table 5, 4.74% error in average)	Requirement of vehicle trajectory data, obtained by short-term WIM-Vision data fusion monitoring.More parameters to calibrate (Eq. 3, Table 4)