Advanced Control Strategy Enhances Stability of Distributed-Drive Electric Vehicles
In an era where electric mobility is rapidly reshaping the automotive landscape, ensuring vehicle stability—especially under extreme driving conditions—has become a critical engineering priority. A groundbreaking study published in the Journal of Chongqing University of Technology (Natural Science) introduces a novel hierarchical control architecture that significantly improves the lateral stability of distributed-drive electric vehicles (DDEVs) by intelligently coordinating direct yaw moment control (DYC) and all-wheel steering (AWS). The research, led by Professor Changgao Xia and graduate student Yazhou Li from the School of Automotive and Transportation Engineering at Jiangsu University, presents a robust, real-time capable strategy that leverages phase-plane analysis, dual sliding-mode control, and optimal tire force allocation to maintain vehicle stability even on low-friction surfaces.
Distributed-drive electric vehicles—equipped with individual motors at each wheel—offer unparalleled control flexibility compared to conventional powertrains. However, this architectural advantage also introduces complex dynamics that demand sophisticated coordination between steering and torque distribution systems. Traditional approaches often treat yaw stability and steering control as separate functions, leading to suboptimal performance when both systems operate simultaneously without proper synergy. The new strategy proposed by Xia and Li addresses this gap by integrating AWS and DYC into a unified framework that dynamically adapts based on real-time vehicle state and road conditions.
At the heart of the innovation lies a two-layer control architecture. The upper layer employs a dual sliding-mode controller that tracks two key reference signals derived from a linear two-degree-of-freedom bicycle model: ideal lateral velocity and ideal yaw rate. These references represent the desired vehicle behavior under stable driving conditions. The controller continuously compares actual sensor-measured values with these ideals and generates two primary outputs: an additional steering angle correction and an additional yaw moment demand. Sliding-mode control was chosen for its inherent robustness against parameter uncertainties and external disturbances—a critical feature for real-world driving scenarios involving crosswinds, uneven road surfaces, or sudden maneuvers.
What sets this approach apart is how the lower control layer interprets and executes these high-level commands. Rather than applying corrections blindly, the system uses the ω–β phase plane—plotting yaw rate against sideslip angle—to assess the vehicle’s proximity to instability. This phase-plane method, long used in theoretical vehicle dynamics, is operationalized here as a real-time decision-making tool. Within a mathematically defined stability boundary (expressed as ω + B₁β ≤ B₂), the system relies primarily on steering adjustments. But once the vehicle state crosses this threshold—indicating incipient oversteer or understeer—the controller seamlessly activates DYC to supplement steering with differential torque distribution across the wheels.
This transition is not abrupt but carefully coordinated. As the vehicle approaches the stability limit, the AWS system commands maximum feasible steering angles based on current tire loads and road friction. Any residual stability demand beyond what steering alone can provide is then fulfilled by DYC through precise longitudinal force manipulation at each wheel. This synergy ensures that both systems operate within their optimal domains: steering for fine, high-bandwidth corrections in the linear tire region, and torque vectoring for coarse, high-magnitude interventions in the nonlinear regime.
A key technical achievement of the study is the formulation of the tire force allocation problem as a constrained optimization task. The lower controller minimizes a cost function representing tire utilization—specifically, the sum of squared normalized longitudinal forces across all four wheels—subject to physical constraints imposed by the friction ellipse. This ensures that no tire exceeds its adhesion limit while collectively generating the exact longitudinal force and yaw moment demanded by the upper layer. The solution respects both equality constraints (total force and moment balance) and inequality constraints (tire force limits dictated by vertical load and road friction coefficient). The resulting torque commands are then mapped to individual wheel motors through a simplified yet accurate hub motor model.
To validate their strategy, Xia and Li conducted high-fidelity co-simulations using CarSim and MATLAB/Simulink. They selected the double-lane-change maneuver—a standard test for assessing emergency avoidance capability—at a constant speed of 54 km/h across varying road conditions (μ = 0.3, 0.6, and 0.9). Three control configurations were compared: AWS-only, DYC-only, and the proposed coordinated AWS+DYC strategy. The results were unequivocal. Under all tested conditions, the coordinated approach yielded the closest tracking to the ideal vehicle response in terms of lateral velocity, yaw rate, and lateral displacement. Notably, on the low-friction (μ = 0.3) surface—simulating wet or icy roads—the coordinated controller maintained trajectory accuracy with minimal overshoot or oscillation, whereas the AWS-only system exhibited significant deviation due to insufficient corrective authority, and the DYC-only system, while stabilizing, produced larger steady-state errors.
The performance advantage was particularly evident during the high-dynamic phases of the maneuver. In the AWS-only case, once tire slip angles exceeded the linear range, further steering inputs yielded diminishing returns and even destabilizing effects. In contrast, the coordinated system activated DYC precisely when needed, using differential torque to generate additional yaw moment without increasing lateral tire load. This not only restored stability but also preserved driver intent by minimizing unintended path deviation. The four-wheel steering angles, computed in real time based on load transfer and desired cornering behavior, demonstrated smooth, physically plausible trajectories that respected mechanical limits.
Beyond immediate stability benefits, the proposed architecture offers several practical advantages for real-world implementation. First, it relies only on commonly available sensor data—vehicle speed, yaw rate, lateral acceleration, and steering angle—making it compatible with existing electronic stability control (ESC) hardware. Second, the phase-plane decision logic is computationally lightweight, enabling execution on standard automotive microcontrollers without requiring high-end processing units. Third, the modular design allows for incremental integration: manufacturers could initially deploy the AWS component and later add DYC as motor control capabilities mature.
From a safety perspective, the implications are profound. Loss of vehicle control during cornering remains a leading cause of single-vehicle crashes, especially on rural roads with unexpected surface changes. By proactively managing the transition between linear and nonlinear handling regimes, this control strategy could prevent many such incidents. Moreover, as autonomous driving systems become more prevalent, such robust stability controllers will be essential for ensuring safe operation under degraded conditions—whether due to sensor failure, adverse weather, or unexpected obstacles.
The research also contributes to the broader academic discourse on integrated vehicle dynamics control. While prior studies have explored AFS/DYC coordination or hierarchical tire force allocation, few have combined phase-plane stability assessment with dual sliding-mode control and real-time optimal allocation in a unified framework tailored for DDEVs. The work by Xia and Li bridges theoretical rigor with practical feasibility, offering a template that can be adapted for various vehicle architectures—including four-wheel independent steering and drive (4WIS-4WID) platforms.
Looking ahead, the authors suggest several avenues for future work. These include extending the strategy to handle combined longitudinal and lateral maneuvers (e.g., braking while cornering), incorporating predictive elements using road preview data, and validating the controller on physical test vehicles. Additionally, integrating driver state monitoring could enable personalized stability interventions—more aggressive for performance drivers, more conservative for elderly or novice users.
In summary, this study represents a significant step forward in the quest for safer, more controllable electric vehicles. By harmonizing steering and torque vectoring through intelligent, physics-informed decision-making, the proposed control strategy not only enhances stability but also maximizes the unique capabilities of distributed-drive architectures. As the automotive industry accelerates toward electrification and automation, such innovations will be crucial in building public trust and realizing the full potential of next-generation mobility.
Authors: Changgao Xia, Yazhou Li
Affiliation: School of Automotive and Transportation Engineering, Jiangsu University, Zhenjiang 212013, China
Published in: Journal of Chongqing University of Technology (Natural Science), 2024, Vol. 38, No. 4, pp. 31–38
DOI: 10.3969/j.issn.1674-8425(z).2024.04.005