1. Introduction
Drivers of commercial vehicles face prolonged exposure to harmful vibrations from poor road conditions and extended operating h, leading to chronic physical and mental health risks [
1,
2]. To address this, seat suspension systems, comprising springs, dampers, actuators, or combinations thereof, have emerged as critical solutions, isolating drivers by absorbing and dissipating road-induced vibrations. These systems attenuate vibration transmission, enhancing comfort and safety. Seat suspension systems are categorized as passive, active, or semi-active. Passive systems [
3,
4], though cost-effective and simple, lack adaptability due to fixed damping properties. Active suspension systems [
5] excel in vibration isolation using external power sources but suffer from complexity, high energy use, and cost barriers. Semi-active systems bridge this gap by modulating damping or stiffness in response to driving conditions, offering effective vibration control without requiring substantial external energy, making them reliable and practical for real-world applications[
6,
7]. By balancing performance and efficiency, semi-active suspension systems represent a promising solution to safeguarding driver well-being while maintaining operational feasibility in commercial vehicles.
In recent decades, MR materials have emerged as a key enabling material for semi-active suspension systems [
8,
9,
10]. MR fluids consist of micron-sized ferromagnetic particles suspended in a carrier fluid. Their rheological properties can be rapidly and reversibly altered by applying an external magnetic field [
11,
12,
13]. MR dampers, which utilize MR fluids, offer several advantages, including a wide range of controllable damping forces, fast response times, and low energy consumption [
14]. These characteristics have made MR dampers a popular choice for semi-active seat suspension systems, attracting significant research interest [
15,
16,
17]. McManus et al. [
18] integrated an MR damper into a seat suspension system to mitigate end-stop impacts. The vibration dose value (VDV) was reduced by approx. 40% compared to traditional dampers, demonstrating that the use of an MR damper can result in considerably less severe impacts and correspondingly lower vibration exposure levels. Hiemenz et al. [
19] implemented semi-active MR dampers in helicopter seat suspension systems, reducing vertical vibration by 76% compared to passive seats, highlighting the excellent performance of MR dampers in seat suspension systems.
MR dampers can be categorized into linear dampers [
20] and rotary dampers [
7,
21] according to their motion modes. Linear dampers produce damping force through the linear motion of the piston, making them ideal for linear vibration control. They are characterized by a simple structure, rapid response times, and the capacity for accurate damping force adjustment. Rotary MR dampers produce damping force through rotational motion, making them suitable for rotational vibration control. They feature a compact design and no stroke limitations but have some limitations in terms of control complexity and response speed. A linear MR damper was used in the seat suspension system in this research in order to utilize its fast response and precise control capabilities to improve vibration isolation performance.
The performance of MR-damper-based seat suspension systems heavily depends on the semi-active control strategy employed. Numerous control methods have been explored to optimize vibration isolation and ride comfort. For instance, Yao et al. [
22] proposed a control strategy based on Lyapunov functional theory and linear matrix inequalities, giving rise to a 39.25% reduction in the root mean square value of body vertical acceleration. Shin et al. [
23] introduced an adaptive fuzzy controller. Integrating this controller with a sliding mode controller further enhanced robustness against mass uncertainty and external disturbances. The results demonstrated that the hybrid controller effectively reduced the vertical acceleration of the seat under bump-related, random, and sinusoidal conditions. Additionally, Dong et al. [
24] compared five semi-active control algorithms, i.e., skyhook, hybrid, LQG, sliding mode, and fuzzy logic control. The five control algorithms attenuated sprung mass acceleration by 13.82%, 0.71%, 3.15%, 14.31%, and 4.6%, respectively.
Despite these advancements, many existing studies rely on overly simplified seat suspension models that fail to capture the inherent nonlinearities and uncertainties of real-world suspension systems. Semi-active seat suspension systems are inherently nonlinear, with suspension springs and dampers exhibiting complex behaviors under varying road conditions and vehicle speeds, posing challenges for accurate suspension control modeling. Furthermore, changes in passenger load or dynamic forces introduce parametric uncertainties, leading to modeling inaccuracies and potential instability in control systems. To address these challenges, Astolfi et al. [
25] systematically proposed a backstepping method for nonlinear systems, utilizing recursive Lyapunov-based design and adaptive laws to estimate unknown parameters, ensuring global stability. Pang et al. [
26] further developed a constrained adaptive backstepping controller for nonlinear active suspension systems, mitigating road disturbances and safety constraints. Simulations demonstrated reductions in vertical acceleration by 87.91% (bumpy roads), 85.26% (random roads), and 95.92% (harmonic excitation). Sun et al. [
27] introduced an adaptive backstepping control strategy for active suspension systems subject to hard constraints. Simulation results showed that the vehicle’s vertical acceleration was reduced by up to 98.29%.
However, the time delay of seat suspension systems is not addressed in the previous studies on backstepping control strategies. The time delay in these systems primarily arises from two sources: (1) the inherent response time of the MR fluid and electromagnetic coil in the MR damper, and (2) the delay caused by the time needed for signal processing and computational output when the MR damper operates in conjunction with external control systems (e.g., sensors and controllers). These delays may impair control performance and lead to system instability [
28,
29,
30]. Majdoub et al. [
31] developed dual observers specifically for MR damper time-delay compensation, enabling real-time estimation of hysteretic internal states and stable adaptive state feedback regulation. While enabling high-accuracy tracking control with guaranteed stability, this approach necessitates precise Bouc–Wen model parameterization and consumes additional computational resources for state observation. Consequently, its practical implementation is subject to non-negligible constraints. Jin et al. [
32] adopted radial basis function (RBF) neural networks for approximating the uncertain time-delay function, while they employed Lyapunov–Krasovskii functionals to rigorously analyze the stability of delayed system components. The proposed methodology synthesizes backstepping control, sliding mode control, neural network approximation, and delay compensation schemes, creating a sophisticated theoretical framework that amalgamates cutting-edge control techniques, albeit with increased complexity in terms of controller realization and practical deployment. Gao et al. [
33] introduced a compensated backstepping controller utilizing an adaptive radial basis function (RBF) neural network inverse model. The Smith predictor compensation strategy was employed to resolve the time delay problem. This approach integrates multiple modules, such as the development of an RBF neural network inverse model, PID compensator design, and Smith predictor compensation. Despite its comprehensive functionality, this method necessitates training the neural network and adjusting multiple compensation parameters, leading to increased complexity.
In this research, a combined strategy of backstepping control and time delay integral compensation was employed to tackle the time delay problem. The time delay integral compensator adds an integral term to the error definition (i.e., the difference between the ideal and actual trajectories), enabling the accumulation of control inputs from a previous time interval. By leveraging the cumulative effect of the integral term, the controller can preemptively adjust the control input to mitigate the lag induced by time delay. This approach eliminates the need for intricate model prediction or parameter tuning and guarantees global stability in the presence of time delay and nonlinear disturbances via Lyapunov stability analysis. Consequently, this solution offers a simplified design, robust performance, and improved adaptability for real-time seat suspension control.
Building on these developments, in this study, we propose an adaptive backstepping control strategy with time-delay compensation to address the challenges of damping force control in MR-damper-based semi-active seat suspension systems. The proposed approach accounts for different types of excitations, mass variations, and input delays, ensuring robust and stable performance. The remainder of this paper is organized as follows:
Section 2 presents the experimental setup and modeling of the MR seat suspension system;
Section 3 details the design and simulation of the proposed controller;
Section 4 describes the construction of a seat suspension vibration test system and the experimental validation of the controller; and
Section 5 concludes the study.
3. Design of a Semi-Active Control System
This Section presents the design of a semi-active control system for an MR-damper-based seat suspension system, addressing challenges relating to time delays, parameter uncertainties, and nonlinear hysteresis. The control framework, illustrated in
Figure 10, integrates five core components: the controlled seat suspension system, which models the physical dynamics of the seat; a reference model combining high-pass and low-pass filters to balance ride comfort and stability, where the high-pass filter attenuates high-frequency vibrations critical for ride comfort, and the low-pass filter suppresses low-frequency oscillations essential for stability; an adaptive backstepping controller that dynamically adapts to disturbances, including driver mass variations, through real-time parameter adaptation; a time-delay compensator that mitigates delays arising from signal processing and actuator response; and an MR damper inverse model, which translates the force command from the controller into the precise control current required to achieve the desired damping force, accounting for the hysteresis and velocity-dependent behavior of the damper. Simulations under harmonic, bump-related, and random excitations validate the effectiveness in enhancing vibration isolation and stability across diverse operating conditions. The control process is initiated by feeding the displacement error
(between the actual response
of the seat system and the ideal trajectory
of the reference model) into the virtual controller and ABC-C controller. The ideal velocity reference trajectory
of the virtual controller, combined with the error
between the seat response velocity
and a historical control input integral term
, is fed into the mass parameter estimator
to generate adaptive parameters for the ABC-C controller. The optimal control force
from the ABC-C controller is partially routed to a delay module, where the generated delayed control force signal simulates actuator time delays. Simultaneously, the control force is input into an inverse MR damper model, converting it into control current
applied to the MR damper for seat suspension vibration control.
3.1. Reference Model Design
This Section details the design of a reference model for generating ideal trajectories for driver mass displacement and vertical velocity, ensuring a balance between ride comfort and stability. The reference model incorporates a performance function that adapts to road disturbances while maintaining stability through Lyapunov analysis.
The reference state vector
is defined as
. Then, the filter performance function is adopted. The performance function
is formulated as follows [
36]:
The transfer functions of these filters are defined as follows:
where
represents a high-pass filter, while
denotes a low-pass filter.
and
are nonlinear functions used to adapt the cutoff frequency of the filters according to the relative displacement of the seat suspension system. They are defined as follows:
where
and
are positive constants.
and
are the design parameters of the high-pass filter and low-pass filter, respectively.
,
,
, and
are positive constants satisfying
. When the relative displacement of seat suspension is large, the cutoff frequency of the high-pass filter increases to minimize the transmission of low-frequency vibrations, improving comfort, while the cutoff frequency of the low-pass filter decreases to reduce high-frequency vibration transmission, enhancing stability. When the relative displacement of seat suspension is small, the cutoff frequency of the high-pass filter decreases to permit more low-frequency vibrations, improving stability, while the cutoff frequency of the low-pass filter increases to allow more high-frequency vibrations, enhancing comfort.
Next, the backstepping adaptive control technique is adopted to minimize
. The second performance function
is defined as follows:
The expected velocity of the reference model
can be calculated as follows:
where
is a positive constant.
,
,
, and
.
Considering the Lyapunov functional candidate below,
by taking the derivative of Equation (17), one can obtain
where
The control input function
is designed in the following manner:
where
is a positive constant. Substituting (20) into (19) yields, Equation (19) can be rewritten as
Then, the Lyapunov functional candidate can be defined as follows:
Since is positive-definite while is negative, and will achieve global asymptotic stability; i.e., when , , . In summary, the reference model inputs the designed control force into the seat suspension system to generate ideal displacement and velocity reference trajectories. By minimizing the errors between the displacement and velocity of the seat suspension system under ABC-C control and the ideal reference trajectories, excellent control performance can be achieved.
3.2. Adaptive Backstepping Tracking Controller Design
This Section presents the design and stability analysis of an adaptive backstepping controller with a compensator (ABC-C) for a semi-active seat suspension system. The ABC-C controller ensures precise tracking of the trajectories of the reference model while compensating for time delays and parameter uncertainties (e.g., driver mass variations). The design process is structured into two phases: (1) the derivation of the control input and adaptive laws and the introduction of a time delay compensator, and (2) a Lyapunov-based stability analysis.
In the first step, the objective is to derive the control input
and the adaptive control law
. Time-delay is unavoidable in a real-world control system, so the control capacity of the controlled seat suspension system will be affected. At this point, compensation is needed. The tracking error between the vertical displacement of the seat system
and the ideal vertical displacement of the reference model
is defined as
. If
tends to be zero,
can track the reference trajectory
more accurately. The vertical motion of the control object is defined below:
where the uncertain parameter
and
.
By taking the derivative of
, one will find that
Similarly, a compensator is designed by defining the error between the reference trajectory
, the virtual control input
in (24), and the integration of the historical control input
, as shown below:
Substituting Equation (25) into Equation (24) yields
By taking the derivative of
, one arrives at
Next,
is defined as follows:
where
is a positive constant. To achieve the control objectives,
is designed in the following manner:
where
is a positive constant and
is the estimation value of
. The adaptive control law based on the
operator is designed in the following manner [
26]:
where
is the tuning parameter for the adaptive control law and
.
The
operator
is given by the following equation:
In the second step, a stability analysis of the designed controller is conducted to prove the global asymptotic stability of the closed-loop system. Toward this end, the Lyapunov function must be defined as follows:
where
. For any
, a set
is always a compact set:
In a compact set
,
has a maximum value. Given that
is continuous,
also has a maximum value. In addition, considering the perfect nature of the square, it can be expressed as follows:
After taking the derivative of
and substituting it into Equations (27)–(31), one obtains
where
. As can be seen from Equation (35), if
it can be formulated as
to obtain
According to Equation (38), when
,
,
, and
have a boundary. By inputting the ideal control force computed through the semi-active controller into the inverse model of the MR damper, the control current, and subsequently the semi-active control law, can thus be determined as follows:
where
is the calculated control current, and
is the maximum current output of the MR damper, which is 2A.
3.3. Numerical Simulation Results and Analysis
Numerical simulations were conducted in MATLAB/Simulink to evaluate the performance of the MR seat suspension system under harmonic, bumpy, and random road excitation conditions. The following controller parameters were selected: , , , , and . The reference model parameters were set to , , , , , , , , and . Four control strategies were compared to validate the effectiveness of damping variability:
- (1)
Passive control—conventional seat suspension with constant damping, with A.
- (2)
Skyhook control—classical on–off logic is used to regulate damping forces.
- (3)
Adaptive backstepping control (ABC)—control forces are adjusted in real-time based on passenger mass variations, but time delays are neglected.
- (4)
ABC-C control—ABC is enhanced by incorporating a time-delay compensator for improved system stability.
3.3.1. Simulation of Harmonic Excitation
A frequency sweep test was conducted on the seat suspension system, and the results are shown in
Figure 11.
Figure 11a shows the sweep excitation with a displacement of ±8 mm and a frequency range of 0.1–5 Hz.
Figure 11b presents the excitation acceleration and response (for the seat’s upper plate) acceleration.
Figure 11c shows the results of a frequency domain analysis of the response acceleration, revealing that the seat suspension system’s resonance frequency is approximately 2 Hz. Therefore, a 2 Hz harmonic excitation was selected for simulation and testing.
The results of the simulation under harmonic excitation conditions are presented in
Figure 12. As shown in
Figure 12a,b, the acceleration and relative displacement amplitude under ABC-C control were significantly reduced, by 67.21% and 63.11%, respectively, compared to the passive case. This is a good trade-off between ride comfort and stability. To illustrate the time-delay phenomenon, the peak acceleration moment at
t = 2.22 s of the passive suspension was selected for analysis (marked by the red dashed box in
Figure 12a).
Figure 12d displays a magnified view of the 2.18–2.28 s time interval with the time delay explicitly marked. The results show that at the peak acceleration moment (
t = 2.22 s), the ABC-C control current had already reached its maximum in advance, while the ABC control current exhibited a 40 ms lag and was still in its rising phase, indicating significant lag in ABC control and demonstrating that ABC-C control can achieve advanced current application through time-delay compensation. Compared to ABC control, the acceleration and displacement amplitudes of ABC-C control are 20.67% and 31.93% lower, respectively. This improvement is attributed to the time-delay compensation mechanism in the ABC-C control system.
3.3.2. Simulation of Bump Excitation
Impact excitation simulations were performed to assess the instantaneous dynamic response characteristics of the seat suspension system. The simulated excitation signal shown in
Figure 13 was derived from experimentally acquired vehicle speed bump traversal data. The figure shows that the 1.38–1.78 s period corresponds to the vehicle crossing the speed bump. The simulation results are presented in
Figure 14.
Figure 14a,b demonstrate that ABC-C control exhibits significantly smaller oscillation amplitudes than passive control. The current command profiles in
Figure 14c reveal a fixed 2A current output of skyhook control before and after crossing the speed bump. During these periods, the seat response velocity aligns with relative velocity direction, requiring maximum damping force output for stability. When crossing the speed bump, skyhook control employs switching strategies to rapidly adjust current based on seat state changes for vibration attenuation. The ABC and ABC-C control systems generate larger and denser current outputs during bump crossing to stabilize the seat while maintaining lower current levels otherwise. The peak-to-peak data in
Table 3 confirm that ABC-C control demonstrates superior performance compared to all the other control systems. Compared with passive control, acceleration decreased by 37.54%, and relative displacement reduced by 27.22%. Compared with ABC control, acceleration and relative displacement decreased by 8.72% and 10.99%, respectively.
3.3.3. Simulation of Random Road Excitation
The comprehensive performance of the seat suspension system was finally evaluated through a random excitation simulation. We adopted ISO 7096 [
37] to generate random road excitation (
Figure 15a), which is specifically applicable to seat vibration testing for engineering vehicles and can effectively evaluate seat suspension damping performance.
Figure 15b,c illustrate the time-domain analysis results for seat acceleration and relative displacement. The results demonstrate that the skyhook, ABC, and ABC-C control systems effectively reduce vibrations, with ABC-C control achieving the best vibration suppression.
Figure 15d displays the command current. In contrast to skyhook, the command current of the ABC-C control system peaks when there are large vibration amplitudes in order to deliver adequate damping force for vibration reduction, but it remains at a lower level when there are small vibration amplitudes. This indicates that the applied current should be adjusted based on the intensity of the vibration. The frequency-domain results regarding seat acceleration in
Figure 15e reveal that the dominant vibration frequency is 2 Hz. Compared to the passive seat suspension system, ABC-C control decreases the seat acceleration amplitude at this frequency by 62%, further demonstrating its superior performance in comparison to all the other control strategies.
We utilized the ISO 2631-1 [
38] standard to assess seat suspension comfort, wherein the frequency-weighted root mean square (FW-RMS) of seat acceleration was used to evaluate the ride comfort. Furthermore, the VDV, a metric more sensitive to peak acceleration, was incorporated into the evaluation framework. Additionally, the root mean square value of the relative displacement (RMS-rd) of the seat suspension system was used as a stability indicator to assess its performance. The evaluation was based on three metrics, namely, FW-RMS, VDV, and RMS-rd, for which lower values signify superior vibration damping performance of the seat suspension system. The results depicted in
Figure 16 indicate that the ABC-C control system, denoted by the red bar, yielded the smallest magnitudes for all the evaluation metrics. In comparison to the passive seat suspension system, the ABC-C control system had a 48.6% lower FW-RMS, a 62.3% lower VDV, and a 56.69% lower RMS-rd. Compared to ABC control, ABC-C control achieved a 31.52% lower FW-RMS, a 35.14% lower VDV, and a 20.5% lower RMS-rd. The results conclusively show that ABC-C, through time delay compensation, significantly enhanced the vibration performance of the seat suspension system, markedly reducing both seat acceleration and relative displacement, thereby effectively balancing ride comfort and stability.
To study the impact of load mass variation on the seat suspension system, simulations were conducted with four different masses under the three signal types. The adaptability of the ABC-C controller to mass changes was validated through comparison with passive control. The results are recorded in
Table 4. It was observed that, in passive seat suspension systems, as the mass increases, seat acceleration decreases, potentially causing vehicle instability. Under ABC-C control, the seat acceleration remains relatively constant as the mass increases. This demonstrates that the adaptive backstepping method can adjust the control force output based on load mass changes, enabling the system to adapt to external disturbances and maintain efficient and robust vibration control performance.