A Review of Recent Advances in Roll Stability Control in On-Road and Off-Road Vehicles

Abstract

Despite significant advancements in roll stability control for individual vehicle types, comparative research across on-road and off-road vehicles remains limited, hindering cross-disciplinary innovation. This study bridges this gap by systematically analyzing roll stability control in both vehicle categories, focusing on theoretical foundations, key technologies, and experimental validation methods. On-road vehicles rely on mature technologies like active suspension, braking, and steering, which enhance safety through sensor monitoring, rollover prediction, and integrated stability control. Validation is primarily performed through hardware-in-the-loop simulations and on-road testing. Off-road vehicles, operating in more complex environments with dynamic load changes and rugged terrain, emphasize adaptive leveling, direct torque control, and active steering. Their stability control strategies must also account for terrain irregularities, real-time load shifts, and extreme slopes, validated through scaled-model tests and field trials. Comparative analysis reveals that while both vehicle types face similar challenges, their control strategies differ significantly: on-road vehicles focus on handling and high-speed stability, while off-road vehicles require more robust, adaptive mechanisms to manage environmental uncertainties. Future research should explore multi-system collaborative control, such as integrating active suspension with intelligent terrain perception, to improve adaptability and robustness across both vehicle categories. Furthermore, the integration of machine learning and advanced predictive algorithms promises to enhance the intelligence and versatility of roll stability control systems.

1. Introduction

Road vehicles and off-road vehicles, as the two major categories of modern transportation and working machinery, play vital roles in various application scenarios. Road vehicles (such as passenger cars, commercial trucks, and buses) are primarily designed for paved roads, with performance requirements focused on ride comfort, handling stability, and safety [1,2,3,4]. In contrast, off-road vehicles (such as agricultural machinery, construction vehicles, and all-terrain vehicles) operate in unstructured environments, where their operational conditions are more complex, demanding higher levels of adaptability, robustness, and maneuverability [5,6,7,8,9]. Despite significant differences in their operating environments and technical requirements, rollover stability remains a critical safety control metric for both vehicle types. For road vehicles, rollover stability directly impacts safety during high-speed cornering and emergency obstacle avoidance maneuvers [10,11]. For off-road vehicles, the influence of dynamic load distribution and terrain irregularities makes rollover control a core technology to ensure safe operation in rugged environments [12,13].

Vehicle rollover accidents are a critical safety issue affecting both road and off-road vehicles, with high global incidence and fatality rates. According to statistics from the National Highway Traffic Safety Administration (NHTSA), rollover accidents account for only 3% of all traffic accidents, yet they result in over 30% of passenger vehicle fatalities. This risk is particularly elevated in high-center-of-gravity vehicles, such as SUVs and light trucks, where the likelihood of rollover significantly increases [14,15]. A report by the European Road Safety Council highlights that rollover accidents account for as much as 20% of heavy commercial vehicle crashes, often occurring during high-speed cornering or emergency obstacle avoidance maneuvers. In the off-road vehicle sector, agricultural machinery and construction vehicles also experience a high frequency of rollover accidents due to unstable terrain and load transfer [16]. Data from the Australian Agricultural Safety Foundation indicate that approximately 40% of fatal accidents involving agricultural machinery are related to rollovers [17]. These statistics underscore the critical importance of enhancing roll stability control (RSC) to improve the safety of both road and off-road vehicles.

The roll stability control technology of road vehicles has undergone a profound evolution, transitioning from passive designs to active control systems. The core objective is to enhance vehicle stability and safety during high-speed driving, emergency obstacle avoidance, and complex maneuvering environments [18,19,20,21]. Traditional passive measures, such as optimizing the vehicle center of gravity, increasing track width, and utilizing anti-roll bars, provide limited inherent stability and struggle to adapt to complex dynamic conditions. To overcome the limitations of passive measures, active roll stability control technologies have been extensively researched and applied. The Electronic Stability Program (ESP), as one of the fundamental active safety systems, effectively mitigates rollover risk by applying independent braking forces to individual wheels, especially during high-speed steering or emergency lane changes [22]. Active suspension systems, such as magnetorheological dampers (MR dampers) and hydraulic actuators, adjust suspension stiffness and damping characteristics based on real-time driving conditions, dynamically suppressing roll angles and improving vehicle stability. Furthermore, active steering and torque vectoring control technologies enhance vehicle stability under extreme conditions by optimizing wheel angles and drive force distribution [23]. In recent years, integrated stability control methods based on model predictive control (MPC) have become a research focus. The MPC framework allows for the simultaneous consideration of roll stability, yaw control, and trajectory tracking, enabling more precise dynamic adjustments [4,24,25]. For instance, Wu et al. [11] addressed the conflict between path tracking and roll stability control during emergency avoidance maneuvers by proposing a cooperative control method based on closed-loop feedback Nash game theory. Simulation and hardware-in-the-loop experiments showed that this controller improved path tracking accuracy by 23% and roll stability by 35%, effectively solving the coordination optimization problem of the two control systems. Liang et al. [26] proposed a yaw–roll coordination control strategy based on dynamic safety demands for heavy vehicles in complex road conditions, using a dynamic weight model predictive control approach based on stochastic integrated double Q-learning. Simulation and experimental results demonstrated that this strategy could adjust control weights in real time, improving vehicle stability by 32% in dynamic environments and effectively addressing various instability risks. However, despite the advancements in existing technologies, several challenges remain, including uncertainty in rollover risk prediction, the robustness of control systems under complex conditions, and the impact of actuator response delays on control performance. Moreover, with the development of new energy and intelligent vehicles, traditional control methods need to be deeply integrated with intelligent chassis and motor drive systems to meet the evolving needs of future vehicles. Therefore, a systematic review and analysis of the current theories and methods of roll stability control for road vehicles is necessary to identify the limitations of existing technologies and explore future development directions.

The development of roll stability control technologies for off-road vehicles exhibits significant differences from that for road vehicles, primarily due to the challenges posed by complex terrain, dynamic load variations, and diverse operational modes [27,28,29]. In contrast to road vehicles, which focus primarily on rollover risks during high-speed driving, off-road vehicles (such as agricultural machinery, construction vehicles, and mining transport equipment) face multiple uncertainties, including rugged terrain, uneven load distribution, and steep slope operations. These factors limit the effectiveness of traditional passive rollover prevention measures in practical applications [30,31,32]. Furthermore, with the widespread adoption of electric drive technologies in off-road vehicles [33,34,35,36], new challenges have emerged in the roll stability control of new-energy agricultural machinery. In recent years, as intelligent perception and active control technologies have evolved, the roll stability control of off-road vehicles has increasingly shifted towards adaptive, predictive, and intelligent systems. Firstly, breakthroughs in real-time sensing technologies have laid the foundation for roll stability control in off-road vehicles. High-precision inertial measurement units (IMUs), LiDAR, GPS, and terrain sensors are widely used to monitor vehicle posture, tilt angles, and terrain undulations in real time. Combined with ground contact force estimation techniques, these sensors enable accurate evaluation of rollover risks [37,38,39,40,41]. Active steering and active suspension systems have also been gradually applied to off-road vehicles, enabling real-time adjustments to actuator outputs based on vehicle posture and terrain information, in order to actively suppress roll risks [42]. Moreover, cooperative control strategies related to rollover prevention have been widely implemented in off-road vehicles, facilitating proactive optimization of vehicle posture [43]. For example, Wang et al. [44] addressed the issue of coordinated path tracking and rollover prevention control for autonomous mining trucks operating in complex terrain. A topology-based coordination control strategy was proposed, and Trucksim/Simulink joint simulations demonstrated that the strategy optimized the path tracking error, yaw rate, and lateral offset angle by 45%, 32.5%, and 20%, respectively, significantly improving path accuracy while ensuring driving safety. Chen et al. [45] proposed a robust model predictive control method based on linear matrix inequalities for addressing the instability of electric mobile platforms in facility agriculture under rotary tiller lifting conditions. Experimental results showed that the controller reduced the average absolute errors in yaw rate and lateral offset angle by 46.4% and 38.2%, respectively, effectively improving the steering stability of the electric mobile platform in rotary tiller lifting conditions. Despite the significant progress made in roll stability control technologies for off-road vehicles, several challenges remain. Firstly, the diversity and uncertainty of unstructured terrain demand higher robustness from control algorithms [46]. Secondly, the high energy consumption characteristics of off-road vehicles make energy efficiency a key constraint for technology adoption. Lastly, cost-effectiveness and general applicability are crucial considerations for the commercialization of these technologies [47,48,49]. Given these challenges, there is an urgent need for a systematic review of the current theories and methods for roll stability control in off-road vehicles, in order to identify key challenges and provide theoretical support for the development of future technologies.

Although numerous studies have summarized and analyzed roll stability control for road and off-road vehicles, the existing literature is mostly limited to a single vehicle category, lacking systematic comparative research. As a result, the commonalities and differences in the rollover stability mechanisms, key technologies, and control strategies between these two vehicle categories remain insufficiently explored. To fill this research gap, as shown in Figure 1, this paper provides a comprehensive review of the theoretical methods, key technologies, and experimental validation techniques for roll stability control in both road and off-road vehicles. It conducts an in-depth comparison of their technological developments and discusses future research directions. The structure of this paper is as follows: Section 2 discusses the research methodology. Section 3 introduces the theoretical foundations of roll stability control, including vehicle roll dynamic modeling, rollover prediction indicators, and the basic principles of roll stability control strategies. Section 4 reviews roll stability control technologies for road vehicles, covering roll state estimation, rollover warning, active roll stability control strategies, and related experimental validations. Section 5 summarizes roll stability control methods for off-road vehicles, addressing their unique roll characteristics, state monitoring techniques, rollover warning, active control technologies, and developments in experimental validation. Section 6 presents a comparative analysis of the roll stability control strategies for road and off-road vehicles, summarizing their commonalities and differences and discussing future development trends. Finally, Section 7 concludes the paper.

2. Research Method

This study adopts a systematic literature review approach, consisting of three steps—retrieval, screening, and evaluation—to systematically reveal the knowledge framework and mainstream research perspectives in the field of vehicle roll stability control. First, the relevant literature was retrieved from major academic databases, such as Web of Science, IEEE Xplore, ScienceDirect, Scopus, and Sage, to ensure the broadness and authority of the sources. During the retrieval process, core keywords, such as “Roll Stability Control”, “Vehicle Rollover Prevention”, “Lateral Stability Control”, “Active Roll Control”, and “Anti-Roll System”, were combined with vehicle-type-related keywords, including “Road Vehicle”, “Highway Vehicle”, “Off-Road Vehicle”, “Heavy-Duty Vehicle”, “Mining Truck”, and “Agricultural Vehicle”. A preliminary selection of approximately 600 papers published between 2000 and 2025 was made, with duplicate papers removed. Subsequently, a second round of screening was conducted, focusing on topics such as roll stability control, rollover prevention strategies, and stability enhancement technologies for road and off-road vehicles. The selected papers were then evaluated in depth to identify representative research outcomes and technological evolution pathways. Finally, a subset of the most representative papers was analyzed to summarize the current state of technological development in roll stability control and to predict future research trends.

Figure 2a shows the annual publication trend of research on vehicle roll stability control. It is evident that research in this field has significantly increased over the past decade, likely driven by advancements in control technologies and the growing demand for safety, reflecting scholars’ continuous focus on roll safety and rollover prevention. Additionally, Figure 2b illustrates the keyword co-occurrence network in the field of RSC. It is clear that rollover prevention research is highly correlated with topics such as rollover risk estimation, active suspension, yaw motion control, and MPC. This indicates that achieving comprehensive improvements in vehicle roll safety and dynamic performance relies on multi-system collaboration and integrated control strategies, aimed at enhancing vehicle stability under complex operating conditions.

3. Theoretical Foundations of Roll Stability Control

3.1. Roll Dynamic Model

Research on vehicle roll stability control is highly dependent on precise dynamic modeling to characterize the vehicle roll behavior under different operating conditions [52,53]. Developing an accurate and widely applicable full-vehicle dynamic model not only aids in developing a deeper understanding of roll and rollover mechanisms, but also provides a theoretical foundation for the design and validation of control strategies. Therefore, this section will focus on the process of constructing a vehicle dynamic model, presenting its basic equations, and reviewing the advancements in precise modeling in existing research.

During vehicle operation, multiple forces act on the vehicle, including gravity, the tire–ground interaction force, aerodynamic forces, and the supporting forces from the suspension system. The interaction of these forces determines the vehicle roll response. Particularly under high-speed cornering, emergency evasive maneuvers, or complex road conditions, factors such as changes in the center of mass height, increased tire lateral forces, or suspension deformation can lead to significant roll, and may even cause the tires to lose traction, resulting in rollover [54,55,56]. Therefore, the primary goal of dynamic modeling is to accurately describe the motion states of the entire vehicle and characterize the coupling relationships between the longitudinal, lateral, yaw, and roll motions, while also considering key influencing factors, such as suspension characteristics, tire mechanics, and changes in the center of mass. This improves the model applicability to various operating conditions, and provides the theoretical basis for deriving and applying linear and nonlinear control methods, thus supporting effective roll stability control design [57,58].

To meet these objectives, full-vehicle dynamic models typically adopt multi-degree-of-freedom (DOF) system modeling approaches, where the roll motion is primarily described by the rotational motion equation around the longitudinal axis of the vehicle. This is then integrated with the vehicle yaw, lateral, and longitudinal dynamics for a comprehensive analysis. The model complexity can be adjusted according to research needs, ranging from simplified models (e.g., 2-DOF roll models) to higher-dimensional models (e.g., 14-DOF vehicle models). Figure 3 presents a common vehicle roll dynamic model, with its specific equations derived as follows [14]:

Equations of motion in the longitudinal plane:

$$m\left({\dot{v}}_x-v_y{\omega }_z\right)=\left(F_{\mathrm{xlf}}+F_{\mathrm{xrf}}\right)\cos {\delta }_{\mathrm{f}}-\left(F_{\mathrm{ylf}}+F_{\mathrm{yrf}}\right)\sin {\delta }_{\mathrm{f}}+F_{\mathrm{xlr}}+F_{\mathrm{xrr}}$$

(1)

$$m\left({\dot{v}}_y-v_x{\omega }_z\right)=\left(F_{\mathrm{xlf}}+F_{\mathrm{xrf}}\right)\sin {\delta }_f+\left(F_{\mathrm{ylf}}+F_{\mathrm{yrf}}\right)\cos {\delta }_{\mathrm{f}}+F_{\mathrm{ylr}}+F_{\mathrm{yrr}}$$

(2)

$$\begin{array}{ll}{I}_z{\dot{\omega }}_z=& \mathrm{a}\left[\left(F_{\mathrm{xlf}}+F_{\mathrm{xrf}}\right)\sin {\delta }_{\mathrm{f}}+\left(F_{\mathrm{ylf}}+F_{\mathrm{yrf}}\right)\cos {\delta }_f\right]-\mathrm{b}\left(F_{\mathrm{ylr}}+F_{\mathrm{yrr}}\right)\\ & +0.5B_{\mathrm{f}}\left[\left(F_{\mathrm{xlf}}-F_{\mathrm{xrf}}\right)\cos {\delta }_{\mathrm{f}}+\left(F_{\mathrm{ylf}}-F_{\mathrm{yrf}}\right)\sin {\delta }_{\mathrm{f}}\right]+0.5B_{\mathrm{r}}\left(F_{\mathrm{xlr}}-F_{\mathrm{xrr}}\right)\end{array}$$

(3)

Roll motion equation:

$$I_x\ddot{\phi }=m{h}_x\left({\dot{v}}_y+v_x{\omega }_z\right)\cos \phi +mg{h}_x\sin \phi -K_{\phi }\phi -C_{\phi }\dot{\phi }$$

(4)

where v_x and v_y represent the longitudinal and lateral velocities of the vehicle, respectively. ω_z is the yaw angular velocity. m denotes the vehicle mass. F_xlf, F_xrf, F_xlr, and F_xrr are the longitudinal forces at the four wheels, respectively. F_ylf, F_yrf, F_ylr, and F_yrr are the lateral forces at the four wheels, respectively. I_z is the vehicle moment of inertia about the vertical axis. a and b are the distances from the center of mass to the front and rear axles, respectively. B_f and B_r are the front and rear track widths. I_x is the vehicle roll inertia about the longitudinal axis. φ is the roll angle. h_x is the distance from the center of mass to the roll axis. K_φ and C_φ are the suspension roll stiffness and damping coefficients, respectively.

Figure 3. Vehicle roll dynamic model.

Figure 3. Vehicle roll dynamic model.

Numerous studies have developed dynamic models to analyze vehicle rollover behavior and support control strategy design. Li et al. [59] proposed a nonlinear 8-DOF model for ME-wheel off-road vehicles, and introduced a predictive load transfer rate (PLTR) index that improves early rollover risk detection. Yun et al. [60] developed a 7-DOF model to analyze the influence of suspension parameters on vehicle response, supporting simulation-based testing. Tan et al. [61] combined roll dynamics with an unscented Kalman filter to estimate the dynamic center of mass, and proposed a roll angle–roll rate phase-plane method for accurate, real-time rollover evaluation. Zheng et al. [62] built a Human–Vehicle–Road model for tankers, and identified the curvature radius, speed, and load rate as key rollover factors. Zhang et al. [63] developed a complex model with 4D visualization to assess the impact of vehicle height and speed on roll angle and tire forces in high-CG vehicles.

Furthermore, to systematically reveal the roll dynamics characteristics of vehicles, various studies have adopted different modeling methods and analytical perspectives, continuously advancing research into vehicle roll dynamics and rollover behavior.

Table 1. Representative roll dynamic models in recent studies.

References	Modeling Approach	Application	Analysis Perspective	Simulated Motion	Remarks
[64,65]	Nonlinear suspension model with MPC	Active roll stability control	Effects of different loads and steering conditions	Roll, lateral, yaw motion	Developed an MPC-based active anti-roll control strategy to adjust suspension stiffness and damping
[66,67]	Variable CG height model with lateral motion	Adaptive roll stability control	Impact of different loads and road conditions	Roll, lateral, vertical motion	Designed an adaptive roll stability controller based on variable CG height
[61,62,68]	Full-vehicle dynamic model with modified CRI	Rollover risk prediction	Influence of CG height and lateral acceleration	Roll, lateral acceleration	Proposed an improved Critical Roll Index (CRI) for rollover risk prediction
[69,70]	Vehicle stability analysis model with tire parameter estimation	Influence analysis of rollover factors	Effects of tire parameters	Roll, lateral, vertical motion	Verified the effects of tire parameters on rollover stability
[71,72,73]	Multi-body dynamic model	Heavy-duty vehicle rollover prevention	Structural deformation effects	Roll, yaw, lateral motion	Investigated rollover risk under high-speed cornering with multi-body dynamics
[74,75]	Integrated roll–yaw model with neural network	Intelligent roll stability control	Data-driven approach for active roll control	Roll, yaw, lateral motion	Utilized neural networks to enhance active roll stability control

summarizes representative modeling methods and their applications in this field in recent years.

3.2. Rollover Prediction Index

The design and application of vehicle rollover prediction indicators is one of the core tasks in rollover prevention control research. A reasonable rollover prediction indicator can help vehicle safety systems to identify potential rollover risks and take appropriate proactive or passive safety measures. To date, researchers have proposed various rollover prediction indicators to accommodate different vehicle types and operating conditions. These indicators are typically classified based on their physical significance, calculation methods, and applicable scenarios, including roll angle [76], time to rollover (TTR) [77,78], and load transfer ratio (LTR) [79], among others.

The roll angle represents the vehicle’s lateral tilt around its longitudinal axis, and is commonly used in real-time stability control due to its intuitive physical meaning and ease of measurement via IMUs or dynamic models. While widely applied in active suspension and ESP systems, it cannot directly quantify rollover risk, and is sensitive to vehicle design and load conditions. To improve its predictive value, Xiao et al. [76] analyzed roll angle behavior under varying speeds and steering inputs, revealing drastic directional changes in low-friction, high-speed scenarios. Mosconi et al. [80] developed a high-precision roll angle estimation method using an extended Kalman filter and a dual-mass model combined with IMU and steering sensor data.

TTR refers to the time required for the vehicle to reach a critical rollover condition from its current state. It is an important indicator for assessing the dynamic variation of rollover risk. Its calculation is based on the vehicle dynamic model, and uses the current roll angle, roll rate, and lateral acceleration to predict when the vehicle will reach the critical rollover state. The formula for TTR is as follows:

$$TTR={\displaystyle \frac{{\phi }_{\mathrm{crit}}-\phi }{\dot{\phi }}}$$

(5)

TTR has clear physical significance and can directly quantify the rollover risk, making it widely used in autonomous driving systems and intelligent vehicle control, especially when combined with MPC to optimize rollover prevention strategies. However, since TTR calculation relies on an accurate dynamic model, its application is influenced by vehicle parameter errors and sensor noise. Chao et al. [81] proposed an improved rollover time prediction method based on a multi-layer neural network. By developing a vehicle rollover dynamic model that accounts for the battery box structure and mass, they classified the rollover indicators containing hyperparameters into five categories, simplifying the algorithm structure using a multi-layer neural network and real-time calculation of variable-step operational-state parameters. Zhu et al. [78] introduced a TTR early-warning algorithm based on a BP neural network, which corrects the TTR indicator online by integrating multiple state parameters, such as the roll angle and yaw rate.

LTR measures the imbalance of vertical loads between the left and right wheels of the vehicle. It is one of the key indicators for evaluating vehicle rollover risk, and the calculation formula is as follows:

$$LTR={\displaystyle \frac{F_{\mathrm{zr}}-F_{\mathrm{zl}}}{F_{\mathrm{zr}}+F_{\mathrm{zl}}}}$$

(6)

where F_zr and F_zl represent the vertical load on the right and left wheels, respectively. The LTR value typically ranges from [−1, 1]. An LTR of 0 indicates that the load is evenly distributed between the left and right wheels, and the vehicle is in a stable state. An LTR of ±1 indicates that one side of the vehicle has completely lost contact with the ground, and the vehicle is on the verge of rollover.

The LTR is straightforward to calculate and can directly reflect rollover risk, making it widely used in rollover warning systems for commercial and heavy vehicles. It is also combined with active rollover control to optimize suspension parameters. However, the LTR is significantly affected by sensor accuracy and road conditions, and there is a risk of misjudgment. Lu et al. [82] proposed a method using Kalman filtering to estimate the dynamic LTR in real time, overcoming the limitations of traditional static thresholds and reducing the vehicle stabilization time by approximately 2 s. Sukumaran et al. [83] introduced a rollover warning algorithm for heavy commercial vehicles based on the dynamic LTR. This method computes the LTR value in real time using a three-degree-of-freedom model and compares it with a preset safety threshold to trigger differential braking control.

In addition to the main indicators mentioned above, researchers have also proposed supplementary rollover prediction metrics to enhance the accuracy of rollover risk assessment, as shown in

Table 2. Comparison of other rollover prediction indexes.

Prediction Index	Evaluation Criteria	Equation	Rollover Threshold	Remarks	References
Roll rate ($\dot{\phi }$)	Rate of change in roll angle, used to assess roll dynamics	$\dot{\phi }={\displaystyle \frac{d\phi }{dt}}$	No fixed threshold, usually combined with roll angle	Reflects roll tendency, but requires roll angle for rollover assessment	[84,85]
Lateral acceleration (ay)	Lateral acceleration of vehicle, indicating rollover tendency	$a_y={\displaystyle \frac{v^2}{R}}$	Typically, ay > 0.4 g indicates rollover risk	Suitable for steady-state rollover analysis, but requires suspension considerations	[68]
Roll energy	Kinetic and potential energy of vehicle roll motion	$\begin{array}{l}{E}_r={\displaystyle \frac{1}{2}}{I}_y{\dot{\phi }}^2+\\ mg{h}_{cg}(1-\\ \cos \phi \end{array}$	Vehicle-dependent, no fixed threshold	Useful for predicting rollover trends, but complex to compute	[14,86]
Dynamic stability index (DSI)	Considers multiple factors, such as roll angle, lateral acceleration, and speed	Computed based on multivariable functions	Depends on dynamic modeling, no fixed threshold	Suitable for intelligent vehicle rollover prediction, computationally complex	[80,87]
Tire lateral force ratio (TLFR)	Ratio of tire’s lateral force to its maximum available lateral force	$TLFR={\displaystyle \frac{F_y}{F_{y\max }}}$	TLFR > 1 indicates a limit condition, possibly leading to loss of control	Useful for analyzing tire adhesion limits, but sensitive to load variations	[88]
CG height variation rate (CGVR)	Dynamic variation rate of vehicle’s center of gravity height	$CGVR={\displaystyle \frac{d{h}_{cg}}{dt}}$	Vehicle-dependent, no fixed threshold	Suitable for uneven terrain analysis, computationally intensive	[89,90]

.

3.3. Basic Principles of Roll Stability Control

The core objective of vehicle roll stability control is to suppress the roll angle and reduce the load transfer ratio, thereby preventing rollover accidents under extreme conditions, while ensuring dynamic controllability. This objective typically involves a trade-off between maneuverability and stability, ensuring that the vehicle maintains reliable roll stability during sharp turns, emergency maneuvers, high-speed driving, and in complex terrain, while preserving reasonable handling performance.

In a roll stability control system, key control variables include the roll angle, roll rate, yaw rate, lateral acceleration, and load transfer ratio (LTR). Among these, the roll angle and LTR directly reflect the vehicle rollover risk, while the yaw rate and lateral acceleration are closely related to the overall stability of the vehicle. In the context of active control, the actuated variables primarily involve active suspension stiffness and damping, wheel braking force distribution (differential braking), and active front steering (AFS), which are used to actively adjust the roll dynamics. Active suspension reduces body roll and improves ride comfort by adjusting suspension stiffness and damping forces, but it comes with higher costs and implementation complexity. Active steering enhances maneuverability and stability by adjusting the front/rear wheel angles, making it suitable for integration with Advanced Driver Assistance Systems (ADASs), but it increases system complexity. Differential braking quickly responds to roll risk by distributing braking force without adding extra actuators, but it may reduce vehicle efficiency and accelerate brake wear. Each actuating mechanism has its advantages, and a comprehensive selection or combined control strategy should be employed, depending on the application scenario, to optimize vehicle roll stability.

Model-based control methods are the mainstream strategy for roll stability control, with the core idea being the use of a vehicle dynamic model to construct a control law that adjusts control inputs in real time based on the measurement or estimation of key state variables. Typical roll stability control strategies include the following: Linear Quadratic Regulator (LQR), sliding mode control (SMC), model predictive control, and Fuzzy Logic Control (FLC).

In recent years, researchers have increasingly focused on multi-system collaborative control strategies, which aim to achieve better roll stability by coordinating the control of multiple actuators. The desired outcome of roll stability control is to minimize the rollover risk and enhance vehicle safety without significantly compromising maneuverability. This optimization process requires balancing lateral stability and handling flexibility to ensure good dynamic responses across various operating conditions. For example, during high-speed obstacle avoidance, the control system should maintain reasonable yaw dynamics while ensuring stability, allowing the driver or autonomous driving system to complete the maneuver successfully. Additionally, in special applications such as off-road terrains or mining transport, roll stability control needs to incorporate terrain adaptability to enhance vehicle safety on slopes and complex surfaces. Therefore, roll stability control involves various physical variables and control strategies, with the goal of achieving synergistic optimization of maneuverability and rollover prevention through active adjustments of the vehicle’s dynamic behavior.

4. Roll Stability Control in Road Vehicles

4.1. Roll State Monitoring and Risk Prediction

Roll stability control for road vehicles mainly focuses on special scenarios such as high-speed cornering and emergency obstacle avoidance. To meet both the maneuverability and safety requirements of the vehicle, it is necessary to monitor the vehicle roll motion in real time and issue rollover warning signals in a timely manner, providing the control system with sufficient response time [84,85]. Figure 4 illustrates the typical prediction process of rollover risk [91]. Therefore, accurate state information must be first obtained to assess the vehicle roll dynamics and rollover risk in real time, serving as a crucial input for the control system to enhance roll stability and safety. Sensor-based state estimation is a straightforward and effective method for obtaining roll state variables, including IMUs, accelerometers, and gyroscopes. Numerous studies have focused on optimizing sensor placement, improving measurement accuracy, and reducing signal noise. Rajamani [84], for example, used a combination of lateral accelerometers and gyroscopes to observe key roll parameters (such as roll angle and center of mass height) in real time. The stability of the observer was verified using the Lyapunov indirect method, significantly reducing hardware costs while maintaining accuracy.

Although direct observation methods provide high-precision data, sensor noise often leads to deviations from the true vehicle state, limiting rollover warning accuracy. To address this, many studies have adopted state estimation approaches that fuse dynamic models with sensor data to enhance robustness and accuracy. Wang et al. [92] developed an IMU-based warning system using adaptive extended Kalman filtering and LTR indicators, while Tan et al. [61] proposed a CNN-LSTM method combining UKF and deep learning for early rollover prediction. Hung et al. [72,79] modeled acceleration thresholds using multi-body dynamics, and Zheng et al. [73] enhanced LTR sensitivity with an 18-DOF model by considering suspension and unsprung mass forces. In control applications, Lu et al. [82] designed an adaptive anti-rollover algorithm based on road adhesion identification and sliding mode control. Zhu et al. [78] introduced a neural network–MPC hybrid system using the TTR indicator for coordinated control. For off-road scenarios, Wang et al. [93] developed a warning system for forest fire trucks, validated through simulation and real tests. These studies highlight the diverse advantages of different prediction methods, with future work focusing on adaptive integration based on application needs.

4.2. Active Roll Stability Control Technologies

4.2.1. Single-Actuator Control

When the rollover prediction system detects that a vehicle is at high rollover risk, the active anti-rollover control system immediately intervenes to adjust the vehicle dynamic response and reduce the rollover risk. Currently, anti-rollover control for road vehicles primarily relies on control techniques such as active braking, active steering, and active suspension. These methods optimize vehicle posture and improve stability by adjusting the longitudinal, lateral, and vertical forces at the tires. Extensive research has been conducted on roll stability control using single actuators. Among these, active braking control can influence the vehicle’s yaw moment through brake force distribution (BFD), thereby reducing lateral acceleration and mitigating rollover risk [94]. In addition, differential drive control—by applying torque asymmetrically across the driving wheels—can also generate corrective yaw moments to enhance lateral stability. Both approaches contribute to optimizing the vertical load distribution between the left and right wheels, effectively lowering the vehicle’s rollover propensity. For instance, Zheng et al. [73] developed a hierarchical control system combining sliding mode ABS, differential braking, and slip ratio identification, which reduced the yaw rate and lateral acceleration by 9–18% in J-turn and fishhook tests. Sukumaran et al. [83] integrated rollover prevention and ABS using LTR-based differential braking, effectively maintaining rollover thresholds during extreme maneuvers without extra sensors. Yao et al. [95] designed a control system for three-axle rescue vehicles using fuzzy PID active suspension and MPC-based braking, which showed superior rollover resistance in step and fishhook tests. Addressing braking failure in steer-by-wire vehicles, Zheng et al. [96] proposed a nonlinear LTR and layered control architecture incorporating failure modeling and road adhesion identification, which improved lateral stability by up to 68.18%. Wang et al. [22] proposed an anti-roll control method based on four-wheel independent drive, and designed a sliding mode variable structure controller. Simulation and real-vehicle tests demonstrated that the proposed approach effectively suppressed vehicle roll and yaw motions, thereby enhancing the spatial stability of in-wheel motor electric vehicles on uneven road surfaces. The advantage of active braking and driving lies in their ability to generate a stable yaw moment in a short period, thereby reducing the rollover risk. However, this method may impact the vehicle’s longitudinal performance and increase the thermal load on the braking system.

Active steering control (ASC) adjusts the steering angles of the front or rear wheels to alter the vehicle yaw motion and lateral force distribution, thereby reducing the lateral acceleration and roll moment, effectively mitigating the rollover risk. Compared to active braking control, active steering allows for smoother intervention in vehicle motion without significantly affecting the longitudinal dynamics, optimizing the roll dynamic response and enhancing driving stability. For example, Qi et al. [70] proposed an MPC-based coordinated control strategy combining active front steering (AFS) and direct yaw moment control (DYC), achieving a 30% improvement in stability and a 25% reduction in path error. Botes et al. [97] integrated AFS with differential braking using a nonlinear MPC controller and understeering gradient model, enhancing robustness and improving stability by 32%. Xiao et al. [76] introduced a 3D stability zone based on the lateral speed, yaw rate, and roll angle, improving the average and minimum stability metrics by 23.89% and 50.16% under low-adhesion conditions. Active steering proves to be a precise, non-intrusive method for mitigating roll dynamics through optimal lateral force distribution. However, the accuracy of this method depends on high-performance actuators and precise vehicle dynamic models. In extreme conditions, it may need to be combined with other control strategies, such as active braking or active suspension, to achieve optimal rollover prevention effects.

Active suspension control directly adjusts suspension forces and roll stiffness to reduce body roll angle and optimize load distribution, making it a direct method for rollover prevention [98,99]. Compared to active braking and active steering, active suspension can provide additional control forces through suspension actuators to directly suppress roll dynamics, without significantly affecting vehicle handling and comfort. Therefore, it is considered one of the most direct methods for rollover prevention. In recent years, researchers have conducted in-depth studies on active suspension-based rollover control. For example, Lu et al. [100] developed an adaptive interconnected suspension (AIS-ARS) using hydraulic links and two electromagnetic valves to replace traditional anti-roll bars, reducing costs by 60% and roll angles by 35%. Li et al. [59] combined rollover warning with fuzzy PID to enhance suspension-based rollover prevention, while Wang et al. [64] introduced a wheelbase-preview MPC method that improves ride stability on uneven roads. Ricco et al. [65] proposed an NMPC strategy with adaptive weighting to coordinate the yaw, roll, pitch, and vertical motions, boosting the control accuracy by 40% and steering response by 25%. Pang et al. [101] incorporated road preview into a 9-DOF model predictive controller to enhance comfort and safety on unstructured roads. Iben et al. [102] used SOS-based polynomial tracking to reduce roll angle errors, improving control in semi-active suspension systems. Although active suspension control can dynamically adjust suspension forces and stiffness, this method places high demands on actuator performance and has higher implementation costs. Therefore, in practical applications, it is essential to balance the system complexity, energy consumption, and control effectiveness.

4.2.2. Integrated Stability Control

Although using a single actuator for roll stability control can suppress vehicle roll to some extent, single control strategies often have limitations. Active braking, while capable of quickly adjusting the yaw moment and reducing the rollover risk, can adversely affect the vehicle’s longitudinal motion performance. Active steering can optimize roll dynamics by altering tire lateral forces, but its control ability is constrained by the physical limitations of the front and rear steering angles. Active suspension, though effective in reducing the body roll angle and improving load distribution, incurs high implementation costs and may not provide sufficient lateral stability under extreme conditions.

Under complex dynamic conditions, a single actuator often cannot meet the requirements for vehicle maneuverability and roll stability, leading to the rise of integrated multi-actuator control for rollover prevention [94]. Ataei et al. [103] proposed an integrated yaw–roll predictive control method for all-wheel-drive electric vehicles, coordinating torque distribution and differential braking. This approach prioritized the use of efficient actuators, maintaining the yaw rate, sideslip angle, and roll angle within safe limits, while balancing energy efficiency and safety. Vehicle and simulation tests showed significant improvements in stability on flat and unstructured surfaces, with differential braking only intervening when necessary. Zhou et al. [104] designed a hierarchical control architecture for commercial vehicles, integrating engine torque limitation, differential braking, and active front-wheel steering within an adaptive MPC framework. HIL tests demonstrated effective coordination of actuators under sharp steering, reducing rollover tendencies and enhancing active safety. Her et al. [105] proposed a three-level chassis coordination control system combining differential braking, semi-active suspension, and the active roll moment. Simulations showed that this system significantly improved maneuverability and roll stability compared to single-control approaches. Termous et al. [106] introduced a multi-objective control strategy integrating active steering, differential braking, and active suspension, using higher-order sliding mode and backstepping controllers. This strategy enhanced the handling response and suppressed rollover risks in extreme scenarios. In multi-axle vehicles, Zhang et al. [107] developed an integrated control system using differential braking and active steering with sliding mode control to allocate yaw and lateral forces dynamically. Their system, with optimized subsystems, improved the computational efficiency, robustness, and control performance, increasing the safe driving speed and overall effectiveness.

In conclusion, multi-actuator integrated control can compensate for the limitations of single-actuator systems by coordinating different control strategies, providing superior roll stability control. However, multi-actuator coordination control involves key technical challenges, such as nonlinear coupling, control allocation optimization, and actuator dynamic response matching. Future research could further explore optimization strategies based on intelligent control, improve real-time computational efficiency, and develop actuator fault-tolerant control to enhance the practicality and robustness of multi-actuator rollover prevention systems.

4.3. Experimental Validation of Roll Stability for Road Vehicles

4.3.1. HIL Experiment Tests

HIL testing is an efficient and reliable validation method that is widely used for performance evaluation of active rollover prevention control systems. Through HIL testing, researchers can simulate real-world road conditions in a laboratory environment and optimize control strategies without compromising driving safety [108]. HIL testing relies on virtual–physical interaction technology, coupling actual controllers with high-fidelity vehicle dynamic models in real time, enabling the control strategies to undergo rigorous testing in an environment that closely mimics real-vehicle conditions. A typical HIL testing platform consists of a real-time simulation engine, controller hardware, actuator simulation modules, sensor signal simulation, and data acquisition and analysis systems, providing a comprehensive validation of active rollover prevention control strategies [109]. The real-time simulation engine runs high-precision vehicle dynamic models (such as CarSim, TruckSim, dSPACE, or MATLAB/Simulink) to simulate vehicle responses under various conditions; the controller hardware (e.g., ESC, active suspension, or active steering controllers) executes the rollover prevention strategies; actuator simulation modules replicate the dynamic responses of braking, suspension, or steering systems; sensor signal simulation generates virtual signals for IMUs, wheel speed, steering angle, etc., to match real-world conditions; and the data acquisition and analysis system records test data in real time, evaluating the stability, robustness, and real-time performance of the control strategies.

In recent years, HIL testing has been widely used to assess the effectiveness of roll stability control strategies. As shown in Figure 5a, Zhang et al. [14] built an HIL platform with steer/brake-by-wire and PXI systems, verifying that their TV-NMPC controller maintained the LTR below 0.8 and outperformed traditional NMPC in yaw, sideslip, and roll control. As shown in Figure 5b, Liang et al. [26] used a TruckSim-based HIL setup to validate an integrated steering–suspension controller, which ensured stable control under dynamic conditions. Wang et al. [77] employed dSPACE-based HIL to test cargo-laden vehicles, confirming that a variable-weight controller effectively limited the RI, yaw rate, and sideslip at high speeds and adhesion. Furthermore, as shown in Figure 5c, Wu et al. [11] developed a PXI-based HIL system with a 0.001 s control cycle, where a Nash controller reduced the path error by 15% and improved the roll stability by 20%. Finally, as shown in Figure 5d, Wang et al. [24] constructed a dual-host HIL platform to evaluate steer-by-wire chassis control, which showed superior performance in rollover prevention and stability at speeds of up to 80 km/h under standard testing conditions.

The main advantages of HIL testing lie in its high safety, high repeatability, and low cost, allowing for the iterative optimization of control strategies without relying on real-vehicle testing. Additionally, HIL testing enables researchers to assess controller performance under extreme driving conditions, such as high-speed sharp turns or emergency evasive maneuvers, which is crucial for enhancing vehicle active safety. However, HIL testing still faces certain challenges. Firstly, the requirements for high-precision dynamic modeling and real-time computational capabilities are substantial, as it is essential to ensure that the simulation models accurately reflect the dynamic characteristics of the real vehicle. Secondly, the precise simulation of sensor signals remains a research challenge, as delays and errors in different sensors may affect the reliability of the test results. Moreover, the results obtained from HIL testing still need to be compared with real-world vehicle experiments to validate their feasibility and applicability.

4.3.2. Real-Vehicle Tests

Real-vehicle testing is a critical step in validating the effectiveness of active roll stability control systems in real-world driving environments. The primary objective is to assess the actual impact of control strategies on vehicle roll dynamics, analyze the response time, stability, and robustness of control algorithms under complex conditions, and calibrate the intervention effects of various active actuators. Compared to simulation and HIL testing, real-vehicle testing can reveal interference factors, such as tire nonlinearities, aerodynamic disturbances, and road surface irregularities, which are difficult to replicate in models. This allows control strategies to be more closely aligned with real driving conditions. In recent years, researchers have conducted systematic real-vehicle testing studies on single-actuator control strategies, including active braking, active steering, and active suspension, as well as multi-actuator integrated control strategies.

To verify vehicle dynamic models and control strategies, several studies have conducted thorough real-vehicle and scaled-platform experiments. Zhou et al. [74] validated a lateral dynamic model enhanced by neural networks using real-vehicle data collected via a GNSS/IMU navigation system and Correvit sensor. Tests under random paths and varying speeds showed that adaptive lateral stiffness modeling improved accuracy by compensating for tire nonlinearities while maintaining generalization. Hajiloo et al. [57] tested an MPC-based control strategy on a four-wheel-drive Chevrolet Equinox using dSPACE and high-precision GPS/IMU. During double-lane changes on dry asphalt (μ = 0.9), the system prioritized torque redistribution and activated differential braking only when necessary, effectively reducing intervention frequency and enhancing control efficiency. Zhao et al. [23] validated an integrated rollover control strategy via a three-axis emergency maneuver at 25 km/h. The system, with inertial navigation and active hydraulic actuators, achieved reductions in the lateral load transfer ratio (10.9–14.3%), lateral acceleration (22.8–27.4%), and roll angle (25.2–29.1%), confirming improved stability. For scaled-platform tests, Xiao et al. [53] verified a vehicle model and electro-hydraulic suspension control using a pure electric prototype vehicle. LQR-based control achieved roll angle and suspension deflection reductions of over 25% in double-lane and sinusoidal maneuvers, with simulation errors below 5%. Overall, these experiments demonstrate that diverse rollover prevention strategies can significantly enhance safety while maintaining dynamic performance across various platforms.

5. Roll Stability Control in Off-Road Vehicles

5.1. Roll Characteristics of Off-Road Vehicles

Compared with on-road vehicles, off-road vehicles typically operate in rugged and uneven terrain with significant slope variations, and their driving conditions are further complicated by factors such as load changes, shifts in the center of gravity, and ground irregularities [110,111]. As a result, off-road vehicles are more susceptible to severe rolling or even tipping accidents, which pose significant threats to operational safety and equipment stability [29,112,113]. Among these, agricultural machinery, as a typical off-road vehicle, operates in environments involving complex terrain, soil properties, and dynamic changes in operational loads, leading to highly nonlinear characteristics in the vehicle mass distribution, suspension response, and roll stability [114]. Especially during operations such as hillside work, plowing, or heavy-duty towing, the contact conditions between the agricultural machine and the ground, the dynamic characteristics of the suspension system, and the height of the center of gravity all undergo significant changes, which, in turn, affect the roll stability [115,116,117,118]. In response to these challenges, researchers have proposed various active posture and stability control methods in recent years to enhance the roll stability of agricultural machinery in complex terrains. Specifically, because tractors often use a rigid chassis structure, the vehicle roll angle is mainly absorbed by the suspension and tire deformations, making them more prone to tipping in harsh terrains [119]. The following sections will systematically discuss active roll stability control strategies applicable to agricultural machinery, with tractors as the primary representative.

5.2. Roll State Evaluation and Rollover Warning

Active safety technologies play a crucial role in reducing the rollover risk of agricultural machinery. Figure 6 illustrates the layout of a typical tractor active roll stability control system, which includes sensor observation, state estimation, rollover risk monitoring, and stability control [17]. To achieve effective rollover warning and active rollover control, the system must first acquire key vehicle state information, such as the roll angle, center of gravity height, and LTR. The acquisition of this information relies on sensor measurements and parameter estimation, which provide the foundational data for subsequent control decisions.

Current research on tractor rollover prevention focuses on real-time monitoring and early warning systems, leveraging inertial sensors like accelerometers and gyroscopes to track the roll angle, yaw rate, and other key states during operation on slopes or uneven terrain [39]. Petrović et al. [120] developed a static and dynamic stability analysis algorithm based on mechanical modeling and 3D geometry, which calculates the critical rollover angle under varying speeds and turning radii. While effective for stability evaluation, traditional methods are limited by their reliance on current state variables, hindering their predictive capabilities. To address this, He et al. [51] introduced the Rollover Critical Position (RCP) index and applied exponential terminal sliding mode control (SMC) to enhance response speed. Building on this, Song et al. [121] proposed a rollover stability indicator derived from a time-varying posture model, enabling early risk detection by monitoring the roll angle relative to its critical threshold and activating active steering when necessary. Further advancing system responsiveness, Gao et al. [122] developed a control system that integrates active steering with a 3-DOF rollover dynamic model to adjust the front-wheel angle in real time based on a calculated stability index, effectively reducing the rollover risk. In terms of practical deployment, Liu et al. [123] created an iOS-based app, SafeDriving, capable of real-time rollover monitoring and emergency alerting via onboard or external sensors, showing strong performance across varied tractor types and terrains. Denis et al. [40] proposed a low-cost adaptive observer method using hydraulic pressure data to estimate the lateral load transfer rate; rollover risk is flagged when |LLT| ≥ 0.8. This approach adapts to changes in load, terrain, and center-of-gravity height, offering a robust and affordable solution for rollover risk detection in adjustable agricultural vehicles like harvesters and tractors. Overall, these advances significantly enhance the foresight, adaptability, and practicality of tractor rollover warning systems.

5.3. Active Roll Stability Control Methods

5.3.1. Active Leveling Control

In operations on slopes and complex terrains, agricultural machinery faces significant challenges to roll stability and operational safety, due to dynamic parameter variations and posture instability [124,125,126]. Jang et al. [127] conducted three-dimensional dynamic simulations to analyze the effects of ground slope and obstacle shape on tractor rollover and rearward tipping. The study concluded that slope is the primary factor causing rollover, with angles of up to 50°, while obstacles with a sine shape increase the rollover risk and rectangular obstacles tend to cause rearward tipping. These findings offer valuable insights for tractor safety design.

To improve the stability of agricultural machinery in complex terrains, researchers have developed several active roll stability control methods, utilizing multi-sensor fusion, advanced control algorithms, and coordinated actuator actions. For active leveling control of tractors, as shown in Figure 7a, Jiang et al. [128] designed an active posture control system for four-wheel-drive tractors operating in hilly terrain. The system adjusts the tractor’s roll angle by linking the front and rear axles, and incorporates a slope stability model based on ground support forces. Simulations demonstrated that the system significantly improves lateral rollover stability, offering a practical solution for hilly-area operations. Further enhancing this, Peng et al. [129] proposed a fuzzy switching-gain sliding mode control system for leveling, using a four-point leveling mechanism. Co-simulation results and test bench validations showed that the system can dynamically level the tractor within ±2° on hilly terrain, restoring it to a 0° roll angle, thus providing a high-precision solution for mountain tractor body leveling.

In addition to tractors, active leveling control for other agricultural machinery has garnered significant attention. As shown in Figure 7b, Jiang et al. [13] designed an omnidirectional leveling system to address vehicle body tilting and safety risks during operations on hilly terrain. The system, based on a “three-layer framework” articulated structure, optimizes key parameters using the NSGA-II algorithm. Both simulation and experimental results show that it effectively reduces the maximum tilt angle and leveling time, meeting operational requirements for hilly areas. As shown in Figure 7c, Wang et al. [12] enhanced this system by incorporating a sliding mode synchronization control method based on a disturbance observer, improving the leveling accuracy and response speed. Experimental results indicated a 35.5% reduction in leveling time and hydraulic cylinder synchronization error within ±6 × 10⁻⁴ m, enhancing safety in hilly terrains. As shown in Figure 7d, Hu et al. [130] developed an adaptive leveling system for tracked agricultural machinery, addressing body tilting in complex terrains. The system enables three-dimensional adjustments (height, lateral, and longitudinal) using RecurDyn multi-posture simulation and a hydraulic drive model. Experimental results showed a leveling accuracy of ±0.4°, offering an effective solution for tracked machinery. Additionally, Yang et al. [66] proposed an adaptive anti-rollover control method for articulated vehicles, using a dynamic stability index and predictive model for material weight, significantly improving anti-rollover performance. Combine harvesters frequently operate on uneven and sloped agricultural terrains, where maintaining lateral stability is critical to both operational safety and harvesting performance. To address this, modern harvesters are equipped with active leveling systems that automatically adjust the orientation of the working units—such as threshing and cleaning assemblies—to remain horizontal when the machine is traversing inclines. These tilting mechanisms, often actuated by hydraulic or electromechanical systems, significantly reduce the risk of rollovers and ensure the efficiency of grain separation and collection. Chai et al. [6] studied the dynamic load characteristics of tracked combine harvesters and validated the stability of hydraulic active leveling mechanisms under complex conditions, further enhancing the safety of agricultural machinery operations.

5.3.2. Direct Torque Control

Direct torque control methods, including active suspensions, single-axis momentum flywheels, and single-frame control moment gyroscopes, have been shown to effectively reduce rollover torque in agricultural machinery, enhancing operational safety. In tractor applications, Son et al. [42] improved safety by equipping a tractor with an active suspension system, allowing independent actuator control of height. This system significantly enhanced posture control during omnidirectional rollover tests, effectively preventing rollover and improving stability on rugged terrain. Qin et al. [131] proposed a novel rollover risk control method combining a single-axis momentum flywheel and an active steering system. Their nonlinear dynamic model and stability criteria demonstrated that this system reduced the tilt angle by 40%, enhancing tractor stability on complex terrains. Further advancements were made by Wang et al. [132], who developed an active anti-rollover control method using a single-frame control moment gyroscope to address rollover during steering on slopes, as shown in Figure 8a. Their nonlinear dynamic model, coupled with a SMC gyroscope rotor precession velocity controller, significantly improved stability and reduced the peak tilt angle by 42%, outperforming traditional PID control. Additionally, Ahn et al. [133] introduced a semi-active suspension system utilizing hydraulic-pneumatic mechanisms and an LQG optimal control algorithm. This system, validated using a Kalman filter state observer, effectively reduced the cab’s vertical acceleration by nearly 49%, offering a cost-effective solution for tractor vibration control.

The operational stability of various agricultural machines has drawn increasing attention. Cui et al. [134,135] developed an active suspension control system for sprayer boom machinery, integrating electro-hydraulic servos and multi-sensor fusion to enhance boom attitude control. Tests showed that the system significantly reduced boom tilt oscillations at resonance, with a standard deviation of tilt angle of only 38% of that of the chassis, demonstrating strong terrain adaptability and disturbance rejection. Zhang et al. [136] proposed a dynamic model for counterweight forklifts, analyzing instability using bifurcation theory and introducing a new rollover indicator for steering and obstacle-crossing scenarios. They further designed a staged hydraulic control strategy and adaptive model predictive control system, which improved rollover prevention through coordinated adjustment of the cylinder force and steering ratio. Li et al. [137] developed a hydraulic active control system for trapezoidal sprayer suspensions using a feedforward PID controller. Experimental results showed superior tracking at 0.2 Hz compared to traditional PID, with field tests indicating a chassis tilt of 3.896° and a boom tilt of just 0.453°, confirming effective real-time boom stabilization.

5.3.3. Active Steering

Similarly to the active steering control principles used in road vehicles, some agricultural machinery reduces the rollover moment and prevents tipping by generating lateral forces through tire steering. As shown in Figure 8b, Song et al. [121] addressed the rollover issue in agricultural tractors by establishing a nonlinear time-varying attitude dynamic model using the Lagrangian method. They proposed a rollover prevention control strategy combining SMC and active steering. By designing the SMC algorithm based on dynamic attitude stability indicators and variable structure theory, they successfully transformed the tractor roll and pitch motions into attitude recovery control. Simulation and scale model test results demonstrated that the SMC-AS strategy effectively reduced the tractor’s rollover stabilization time by 46.07%, showing excellent robustness and control performance under various operating conditions. This provides an effective technical solution for preventing tractor rollover accidents. Meanwhile, Gao et al. [122] addressed the issue of fatal rollover accidents in agricultural tractors by developing a low-cost rollover prevention control system based on active steering technology. They established a three-degree-of-freedom rollover dynamic model that incorporates an automatic steering system and proposed a tilt angle adjustment scheme based on adaptive sliding mode control. The system also designed a front-wheel angle tracking controller using internal model control (IMC) theory. Simulation results showed that the system can compute stability indicators in real time and maintain them within a safe range by adjusting the front-wheel angle (with a stability indicator error of <5%), providing a practical solution to reduce tractor rollover accidents.

In addition to tractors, active steering technology for other agricultural machinery has also received widespread attention. Li et al. [38] addressed the issue of the complex installation and vulnerability of traditional steering angle measurement methods in agricultural environments. They proposed a novel dynamic measurement method based on vehicle attitude information and non-contact attitude sensors. Experimental results showed that the proposed method significantly improved measurement reliability without relying on vehicle kinematic models or repeated calibration, effectively avoiding issues such as mud blockage and cable damage. Furthermore, Liu et al. [138] proposed an active steering system for a new four-wheel independently driven agricultural electric vehicle, based on MPC and direct yaw moment control (DYC). Simulation and experimental results demonstrated that when the steering angle was set to 5°, the steering error was below 0.22%, and the yaw moment error was less than 0.17%, verifying that the system could achieve precise autonomous steering and effectively reduce crop crushing in the field.

5.4. Typical Roll Stability Experiments for Off-Road Vehicles

Experimental testing of agricultural machinery, such as tractors, is a critical step in validating the effectiveness of active anti-rollover methods and control strategies [139,140,141,142,143]. Through experimental testing, the performance of different control strategies and anti-rollover technologies under various operating conditions can be accurately assessed, providing theoretical support and technical guidance for practical applications. Scale model testing, as an effective experimental method, is widely used in agricultural machinery research. Its advantages lie in the ability to simulate the behavior of real machinery by reducing the scale of the model, significantly lowering experimental costs and improving repeatability. As shown in Figure 9a, Wang et al. [132] validated the SGCMG system’s rollover protection through a 1:5.5 scale model experiment. The SGCMG system, using sliding mode control, achieved a 0.62 s response time, outperforming PID control. It kept the tilt angle within 40° and demonstrated good adaptability across speeds (0.15–0.38 m/s). As shown in Figure 9b, Song et al. [121] verified the effectiveness of sliding mode active steering control (SMC-AS) in improving tractor stability. SMC-AS reduced the side tilt by 61.3%, shortened the rollover stability time by 46.07%, and extended the safe driving range from 20° to 25°, providing an improvement of 51% in emergency response time.

Compared with the scale model testing, full-scale vehicle testing can more accurately reflect the lateral dynamics of agricultural machinery, further validating the effectiveness of rollover control measures. As shown in Figure 9c, He et al. [51] developed an active rollover prevention control platform for a 15-horsepower small tractor, integrating central control, steering, and navigation units. Full-scale testing on slopes (8–20°) and various off-road surfaces demonstrated that the active steering system could steer the front wheels along the slope, effectively preventing rollover when the risk parameter exceeded the threshold. The system’s effectiveness was validated in complex terrains, with remote start/stop and safety measures ensuring experimental safety. As shown in Figure 9d, Jiang et al. [128] tested the tractor’s lateral stability and attitude adjustment through bench testing. Results showed that active attitude adjustment via the rear axle mechanism increased the tractor’s instability angle from 30.91° to 40.74°, improving lateral stability by 34.26%. The experimental error between the model and results was 8.12–10.14%, mainly due to unmodeled tire deformation. For tracked work machines, as shown in Figure 9e, Jiang et al. [13] verified automatic leveling performance through static and dynamic tests. On a 20° side slope, leveling took 3.3–3.4 s, with an angle of 20.8–21.2°, and at 3 km/h, leveling reduced the tilt angle from 19.2° to 16.4°, stabilizing within ±1.5° after 18.2 s. The experimental results confirmed that the system could effectively maintain the stability of the machine’s tilt angle, meeting performance requirements.

6. Comparative Analysis and Future Directions

6.1. Comparisons of Roll Stability Control Between Road and Off-Road Vehicles

There are significant differences between on-road and off-road vehicles in terms of vehicle roll state monitoring, rollover warning, and stability control. These differences primarily stem from variations in their operating environments, working conditions, and control requirements. Both types use IMUs, gyroscopes, and accelerometers to monitor parameters like the roll angle, lateral acceleration, and load transfer ratio [68,122,144]. However, on-road vehicle state estimation assumes stable tire–road contact and minimal load variation, enabling effective use of Kalman-based filtering methods [61]. In contrast, off-road vehicles operate in rugged, uneven terrains with variable loads and low-frequency vibrations, posing greater challenges. To address these, their systems often integrate terrain slope sensors, LiDAR, and GNSS to enhance perception and accuracy [145,146,147,148]. Additionally, due to frequent changes in the center of gravity—such as when different implements are mounted—off-road vehicles require higher estimation thresholds and adaptive algorithms to maintain monitoring accuracy [149,150].

Rollover warning systems for on-road vehicles are primarily based on the vehicle yaw motion, speed, and steering conditions. Methods such as the load transfer ratio and static stability factor are commonly used to assess rollover risks [76,81]. The ESP typically utilizes real-time data, such as the vehicle speed, steering angle, and lateral acceleration, to predict potential instability situations and intervene when necessary [4,151]. In contrast, rollover warning for off-road vehicles requires the integration of more complex environmental variables. While operating vehicles typically have lower speeds, they may experience significant slopes. Even at static or low speeds, these vehicles may roll over due to uneven terrain or load variations. Therefore, rollover warning systems for off-road vehicles must integrate real-time terrain information, changes in the vehicle’s center of gravity, and tire–ground contact states [39].

In terms of roll stability control, on-road and off-road vehicles employ significantly different control strategies, due to their distinct working environments and dynamic characteristics. On-road vehicles primarily rely on ESPs, active suspension systems, and active steering systems to control roll stability. The control objective of these systems is to improve vehicle stability while maintaining handling performance, and to provide real-time intervention under extreme driving conditions, such as emergency lane changes and high-speed cornering [152]. Off-road vehicles, with their lower speeds, typically face rollover risks primarily due to terrain undulations and load distribution changes. As such, roll stability control for off-road vehicles is more dependent on active steering, center of gravity adjustment, and hydraulic suspension systems [59,136]. For example, tractors often use a hydraulic suspension system with coordinated control to optimize vehicle posture by adjusting the suspension stiffness and tilt angle. Slope-working machines may use an automatic leveling system that adjusts track width or vehicle posture to maintain stability [13]. Additionally, off-road vehicle roll control systems generally require higher robustness and environmental adaptability to handle complex disturbances in extreme conditions. For instance, some advanced agricultural machines have begun to employ electro-hydraulic servo suspension control, integrated with intelligent perception and adaptive control algorithms, to enhance adaptability to complex terrain.

In summary, on-road vehicles focus on real-time stability control under high-dynamic driving conditions, relying on an ESP, active suspension, and steering control to enhance driving maneuverability. Off-road vehicles, however, face more complex terrain and load variations. Their control systems are more reliant on environmental perception and terrain adaptability, employing multi-sensor fusion and active leveling technologies to improve stability and safety under extreme conditions.

6.2. Future Directions

In the future, technologies for roll monitoring, rollover warning, and stability control in both on-road and off-road vehicles should progress towards intelligent, data-driven, multi-sensor fusion, and collaborative control systems to enhance real-time performance, accuracy, and robustness. Despite significant progress, there remain numerous challenges in adapting to complex operating conditions, optimizing computational complexity, and coordinating multi-actuator control. Future research can be focused on the following areas:

• Future roll state monitoring systems will prioritize real-time, robust roll state sensing through advanced multi-sensor fusion (e.g., IMU, LiDAR, GNSS, vision) and adaptive filtering or nonlinear observers to improve accuracy under dynamic and noisy conditions.
• Future systems will apply deep learning and reinforcement learning to build data-driven rollover prediction models, reducing reliance on complex dynamics while improving adaptability to extreme conditions. V2X integration will enable real-time cooperative risk assessment using shared vehicle and road data.
• Control algorithms will increasingly utilize adaptive learning methods, such as deep reinforcement learning and neural networks, to dynamically adjust to diverse driving conditions. Hybrid modeling (physical + data-driven) will enhance robustness and applicability.
• Future systems will evolve toward coordinated control of suspension, braking, and steering, improving stability in high-risk scenarios. Technologies like steer-by-wire and active chassis systems will support real-time structural adjustments.
• To address computational challenges of high-precision controllers (e.g., MPC), research will focus on reduced-order modeling, control parameter tuning via reinforcement learning, and deployment via edge computing to ensure fast and efficient real-time performance.
• For off-road environments, future work will enhance terrain perception through LiDAR, IMU, and vision fusion. Control strategies will include adaptive suspension, variable track width, and real-time center-of-gravity adjustments to maintain stability on uneven terrain.

7. Conclusions

This study presents a comprehensive and comparative investigation into roll stability control for both on-road and off-road vehicles, offering several key contributions to the field. First, by establishing a unified analytical framework, the paper enables a clearer understanding of the fundamental differences and shared challenges in roll stability control across diverse vehicle categories. Second, the study identifies and compares the effectiveness of core technologies—such as active suspension, direct torque control, and adaptive leveling—within distinct operational environments, providing a solid foundation for informed system design. Third, the analysis reveals that while on-road vehicles benefit from high-speed responsiveness and sensor fusion accuracy, off-road vehicles demand enhanced adaptability to environmental uncertainties, prompting a shift toward more robust and terrain-aware control strategies.

A notable contribution of this work is the detailed synthesis of experimental validation approaches tailored to each vehicle type, bridging the methodological gap between simulation-heavy on-road studies and field-based off-road evaluations. Furthermore, the study highlights potential directions for integrating emerging technologies—particularly machine learning, predictive algorithms, and intelligent terrain perception—into roll stability systems, paving the way for more intelligent, adaptive, and universally applicable control solutions.

By clarifying the technological divergences and convergence points between vehicle categories, this research provides a foundation for future cross-domain innovation in vehicle dynamics and control. It not only enhances the theoretical understanding of roll stability, but also contributes practical insights into system implementation under real-world conditions, supporting the development of safer and more adaptable intelligent vehicles.