Adaptive Torque Distribution Enhances EV Cornering Stability
In the rapidly evolving landscape of electric mobility, ensuring vehicle safety during dynamic maneuvers has become a top priority for automotive engineers. As distributed-drive electric vehicles (DDEVs) gain traction for their superior agility and control flexibility, researchers are intensively exploring advanced control strategies to enhance lateral stability—particularly during high-speed cornering on low-friction surfaces. A recent breakthrough from North University of China offers a promising solution through an adaptive torque distribution strategy that significantly improves vehicle handling and safety.
Led by Associate Professor WANG Yanhua and graduate researcher MENG Yongkai at the School of Energy and Power Engineering, the study introduces a novel hierarchical control framework designed to optimize the lateral dynamics of DDEVs. Published in the Journal of Chongqing University of Technology (Natural Science), this research presents a comprehensive approach that combines precise torque calculation with intelligent distribution logic, enabling vehicles to maintain stability under aggressive steering conditions while preserving driver intent as much as possible.
The core challenge in lateral vehicle control lies in managing two critical parameters: yaw rate and sideslip angle. During sharp turns, especially on slippery roads, uncontrolled yaw moments can lead to oversteer or understeer, increasing the risk of skidding or rollover. Traditional stability systems often rely on braking interventions that sacrifice speed to regain control. However, in modern DDEVs equipped with independent in-wheel motors, there is a unique opportunity to leverage torque vectoring not only for stabilization but also for maintaining performance.
WANG Yanhua’s team recognized that a one-size-fits-all control strategy is insufficient for the wide range of driving scenarios encountered in real-world conditions. Instead of relying on a single torque distribution method, they proposed a dynamic, condition-based approach that adapts to the severity of the instability. Their system operates on a two-layer architecture: an upper-level torque calculator and a lower-level torque allocator, both working in concert to deliver optimal performance.
At the heart of the upper controller is a PID-based algorithm that continuously compares actual vehicle behavior with ideal reference values. These references—target yaw rate and sideslip angle—are derived from a two-degree-of-freedom vehicle dynamics model, which takes into account key parameters such as vehicle speed, steering angle, road friction, and mass distribution. The difference between measured and desired values generates error signals that feed into the PID loops, producing both the total drive torque needed to maintain speed and an additional yaw moment required to correct instability.
What sets this strategy apart is how the calculated yaw moment is implemented. Rather than applying a fixed distribution rule, the lower controller selects from three distinct modes—differential drive, differential braking, and friction braking—based on the magnitude of the corrective torque needed and the physical limits of the motors.
In mild cornering situations where only a small yaw correction is required, the system employs differential drive. This method increases torque on the outer wheels while reducing it on the inner ones, generating a stabilizing yaw moment without affecting the total tractive force. Because no braking is involved, vehicle speed remains largely unchanged, allowing the driver to maintain momentum through the turn. This mode is particularly effective on dry or moderately slippery surfaces where traction is still available.
However, when the required corrective torque exceeds what can be achieved through drive modulation alone—such as during sudden evasive maneuvers or on icy roads—the system seamlessly transitions to differential braking. In this mode, the inner wheels are subjected to regenerative or friction braking, creating a stronger yaw moment by decelerating one side of the vehicle. While this results in some speed reduction, it provides a more powerful stabilization effect, crucial for preventing loss of control.
In extreme cases where even differential braking reaches its limit, the system escalates to full friction braking on specific wheels. This highest-intervention mode applies maximum braking force to the inner rear wheel while simultaneously engaging the inner front wheel beyond its motor capacity, effectively using traditional brake actuators to generate additional yaw moment. Although this leads to noticeable deceleration, it ensures that stability is maintained even under the most demanding conditions.
This tiered approach mirrors the philosophy of modern electronic stability programs but leverages the unique capabilities of distributed electric drivetrains. Unlike conventional vehicles where braking is the primary tool for yaw control, DDEVs can use torque vectoring proactively, often avoiding the need for aggressive braking altogether. The adaptive nature of WANG’s strategy ensures that the least intrusive method is always used first, preserving energy efficiency and ride comfort whenever possible.
To validate their control logic, the research team conducted extensive co-simulations using MATLAB/Simulink and CarSim, two industry-standard tools for vehicle dynamics analysis. They tested the system under two challenging scenarios: a sine-steer maneuver on a low-friction surface (μ = 0.3) at 80 km/h, and a double-lane-change maneuver on a high-grip surface (μ = 0.9) at 90 km/h. These tests were designed to simulate emergency avoidance situations commonly used in vehicle safety assessments.
The results were compelling. During the sine-steer test, the uncontrolled vehicle exhibited a peak yaw rate of approximately -14.5 °/s, indicating significant instability. In contrast, the controlled vehicle kept its yaw rate within ±7.5 °/s—a reduction of 54%. Similarly, the sideslip angle, which reached ±3.1° in the baseline case, was suppressed to within ±0.4° under control, closely tracking the ideal zero-reference value. Most impressively, the vehicle maintained its target speed with less than 0.1 km/h deviation, demonstrating that stability and speed retention can coexist under moderate conditions.
The double-lane-change test further highlighted the system’s adaptability. Here, the uncontrolled vehicle showed a peak yaw rate of ±25 °/s, a level that would likely result in loss of control in real-world driving. With only differential drive active, the peak yaw rate dropped by 55%. Adding differential braking improved this to a 57% reduction, and with all three modes engaged, the system achieved a maximum reduction of 58%. This incremental improvement underscores the value of having multiple control tiers available.
Interestingly, the data revealed a trade-off between stability and speed fidelity. When only differential drive was used, the vehicle maintained near-perfect speed tracking. However, when higher intervention modes were activated, speeds dropped to 87.2 km/h and 86.8 km/h respectively, reflecting the inevitable energy dissipation associated with braking. Yet, this sacrifice proved necessary: the enhanced yaw control prevented excessive sideslip and ensured the vehicle stayed on its intended path.
One of the most notable findings was that the full three-mode system did not always produce the best sideslip control. In the double-lane-change scenario, the combination of differential drive and braking performed slightly better than the full system in tracking the ideal sideslip angle. This suggests that over-aggressive correction in the highest mode may introduce minor oscillations, a nuance that could inform future refinements in transition logic.
From an engineering perspective, the success of this strategy hinges on accurate real-time estimation of vehicle states and road conditions. The model assumes knowledge of parameters such as tire-road friction, which can vary unpredictably. While the study used predefined friction coefficients in simulation, real-world deployment would require robust estimation algorithms—possibly integrating sensor fusion techniques using data from wheel speed sensors, IMUs, and camera systems.
Moreover, the control system must account for hardware limitations, including motor torque ceilings and thermal constraints. The paper explicitly models these boundaries, ensuring that commanded torques remain within physically achievable ranges. This attention to practical feasibility enhances the credibility of the approach and increases its potential for real-world implementation.
The implications of this research extend beyond academic interest. As autonomous driving technologies advance, the ability to maintain stability during emergency maneuvers becomes even more critical. Self-driving systems must not only plan safe trajectories but also execute them reliably under all conditions. WANG Yanhua’s adaptive torque distribution strategy provides a robust foundation for such capabilities, particularly in electric platforms where precise torque control is native to the architecture.
Automakers investing in distributed electric drivetrains—such as those developing skateboard platforms or high-performance EVs—could benefit significantly from adopting similar control philosophies. The ability to modulate torque across individual wheels opens up new possibilities for enhancing both safety and driving dynamics. For instance, such systems could be integrated with advanced driver assistance systems (ADAS) to provide preemptive stabilization before a skid occurs, or used in performance modes to tighten cornering lines without sacrificing grip.
Furthermore, the energy efficiency aspect should not be overlooked. By prioritizing drive-based corrections over braking, the system minimizes energy loss and maximizes regenerative braking opportunities. This aligns well with the broader goals of electric mobility: reducing environmental impact while improving performance.
Looking ahead, several avenues for refinement exist. The current PID-based controller, while effective, could potentially be enhanced with model predictive control (MPC) or artificial intelligence techniques to anticipate driver inputs and road conditions. Additionally, incorporating driver behavior models could allow the system to adjust its aggressiveness based on perceived skill level or driving mode preferences.
Another area for exploration is the integration of suspension and steering controls. While this study focuses on torque vectoring, combining it with active steering or adaptive dampers could yield even greater stability improvements. Such holistic vehicle dynamics management represents the next frontier in automotive control systems.
In conclusion, the work by WANG Yanhua, MENG Yongkai, and their colleagues at North University of China represents a significant step forward in the field of electric vehicle stability control. By moving beyond static torque distribution rules and embracing an adaptive, multi-mode strategy, they have demonstrated a practical and effective way to enhance cornering safety without unduly compromising driving experience. Their findings underscore the transformative potential of distributed electric drivetrains—not just as a means of propulsion, but as a platform for intelligent, responsive, and safer mobility.
As the automotive industry continues its electrification journey, research like this will play a crucial role in shaping the next generation of vehicles. It is not merely about replacing internal combustion engines with batteries and motors, but about reimagining what a car can do. With intelligent torque management systems like the one described here, the future of driving promises to be not only cleaner and more efficient, but also safer and more enjoyable.
WANG Yanhua, MENG Yongkai, DENG Jichen, WANG Yaoxun, GUAN Quancai, LIU Yutao, School of Energy and Power Engineering, North University of China. Lateral stability of distributed electric vehicle steering torque adaptive distribution strategy. Journal of Chongqing University of Technology (Natural Science), 2024, 38(3):46-54. doi: 10.3969/j.issn.1674-8425(z).2024.03.005