Smart Torque Control Enhances Stability in High-Speed Electric Vehicles
In the rapidly evolving world of electric mobility, where performance and safety are paramount, a new control strategy is emerging that could redefine how distributed electric vehicles (DEVs) handle high-speed cornering. Researchers from Xi’an University of Architecture and Technology have introduced an advanced torque-based electronic differential control system designed to improve vehicle stability, reduce energy consumption, and enhance driving safety—especially under demanding high-speed conditions.
As the global automotive industry shifts toward electrification, distributed electric vehicles—those equipped with individual hub motors at each wheel—are gaining attention for their superior agility, precise torque control, and mechanical simplicity. Unlike traditional internal combustion engine vehicles or even centralized electric drivetrains, DEVs offer unparalleled flexibility in managing power delivery to each wheel independently. This capability opens the door to sophisticated electronic differential strategies that can significantly influence vehicle dynamics.
However, conventional speed-following electronic differential systems have shown limitations, particularly at higher speeds. These systems typically rely on the Ackermann steering model to calculate ideal wheel speeds during turns. While simple and intuitive, the Ackermann model treats the vehicle as a rigid body and ignores critical nonlinear factors such as tire-road interaction, load transfer, and lateral forces. As a result, these models often fail to maintain stability when pushed to the limit, especially on low-friction surfaces or during aggressive maneuvers like double-lane change tests.
Recognizing these shortcomings, Dr. Nannan Zhao and Bo Shi from the School of Mechanical and Electrical Engineering at Xi’an University of Architecture and Technology have developed a novel two-layer torque distribution control framework. Their work, published in Mechanical Science and Technology for Aerospace Engineering, presents a robust solution that leverages artificial intelligence and optimization theory to deliver superior handling and efficiency.
The core innovation lies in moving away from speed-centric control logic and instead focusing on torque as the primary control variable. This shift allows the system to directly influence the vehicle’s yaw moment—the rotational force around its vertical axis—which is crucial for maintaining directional stability during cornering. By calculating and applying an additional yaw moment in real time, the controller can actively correct deviations in the vehicle’s behavior before they escalate into instability.
At the heart of this strategy is a dual-layer control architecture. The upper layer employs a Genetic Algorithm-optimized Back Propagation (GA-BP) neural network to determine the required corrective yaw torque. Neural networks are powerful tools for modeling complex, nonlinear systems, but they often suffer from slow convergence and sensitivity to initial parameters. To overcome this, the researchers integrated a genetic algorithm—a bio-inspired optimization technique—to pre-tune the network’s weights and thresholds. This hybrid approach accelerates training and improves prediction accuracy, ensuring that the system responds quickly and reliably to dynamic driving conditions.
The lower layer of the controller uses quadratic programming to distribute the total drive torque across the four wheels. This mathematical optimization method ensures that the torque allocation minimizes tire workload while respecting physical limits such as road friction and motor capacity. By keeping the tires operating within their optimal grip range, the system maintains high lateral force reserves, which are essential for avoiding skidding or loss of control.
One of the key advantages of this approach is its ability to account for real-time changes in vehicle dynamics. During cornering, weight shifts from the inside to the outside wheels, altering the vertical load on each tire. The proposed control system continuously monitors these load variations and adjusts torque distribution accordingly. This not only enhances stability but also reduces rolling resistance and overall energy demand.
To validate their strategy, the research team conducted co-simulations using CarSim and Simulink—two industry-standard tools for vehicle dynamics analysis. The test scenario involved a high-speed double-lane change maneuver at 90 km/h on a road surface with a friction coefficient of 0.5, simulating wet pavement conditions. This type of test is widely used to evaluate a vehicle’s transient response and stability control performance.
The results were compelling. Vehicles equipped with the new torque-based control system demonstrated a 6.24% deviation from the ideal trajectory, compared to a 17.47% deviation for those using conventional equal-torque distribution. This means the vehicle stayed much closer to the intended path, reducing the risk of encroaching into adjacent lanes during emergency avoidance maneuvers.
More importantly, the system dramatically improved control over critical stability indicators. The peak sideslip angle—the angle between the vehicle’s direction of travel and its actual orientation—was reduced from 3.35° to just 0.73°, a reduction of nearly 89%. Sideslip angle is a fundamental metric in vehicle dynamics; excessive values indicate that the rear of the car is beginning to slide outward, a precursor to oversteer and potential spin-out. By keeping this angle within ±2.5°, the system ensures the vehicle remains predictable and controllable even under stress.
Similarly, the yaw rate—the speed at which the vehicle rotates around its center—was more closely aligned with the ideal reference model. The controlled vehicle exhibited smoother and more consistent yaw behavior, avoiding the sharp fluctuations seen in the baseline system. This translates to a more stable and confident driving experience, particularly for less experienced drivers who may struggle to correct sudden oversteer or understeer events.
Another significant finding was the system’s impact on energy efficiency. In the simulation, the total energy consumed by the drive motors dropped by 8.2%, from 106,710 joules to 97,930 joules. This reduction stems from two main factors: first, the intelligent torque distribution minimizes unnecessary wheel spin and slippage, which waste energy; second, the system allows for slight speed modulation during turns, reducing aerodynamic and rolling resistance. While 8.2% may seem modest, in the context of electric vehicles where every watt-hour counts, such savings can extend driving range and reduce battery degradation over time.
Moreover, the control strategy enables smoother transitions between straight-line driving and cornering. Instead of maintaining a constant high speed through a turn—which increases lateral forces and destabilizes the chassis—the system proactively reduces speed to a safer level (around 87–88 km/h in the test) and then accelerates back to the target speed after the maneuver is complete. This adaptive speed management is not merely about compliance with physics; it reflects a smarter, more human-like driving style that prioritizes safety without sacrificing performance.
From an engineering perspective, the integration of AI and optimization techniques represents a significant step forward in vehicle control systems. Traditional PID controllers, while reliable, often require extensive tuning and struggle with nonlinearities. Fuzzy logic and sliding mode control have been explored as alternatives, but they come with their own challenges in terms of complexity and chattering. The GA-BP neural network approach offers a data-driven alternative that learns from real-world driving scenarios and adapts to changing conditions without manual recalibration.
The use of CarSim-generated training data from a conventional vehicle with a mechanical differential further strengthens the practical relevance of the model. By mimicking the behavior of a well-understood reference system, the neural network can generalize its learning to a wide range of driving situations, including step inputs, sinusoidal steering, and evasive maneuvers.
Perhaps one of the most underappreciated aspects of this research is its focus on minimizing internal power losses. In all-wheel-drive systems, especially those with independent wheel motors, there is a risk of “circulating power”—a condition where torque is simultaneously applied to both sides of the vehicle, creating opposing forces that cancel each other out and waste energy. The quadratic programming solver explicitly accounts for this by limiting the difference in torque between front and rear axles, thereby improving drivetrain efficiency.
This holistic approach—balancing stability, energy use, and mechanical efficiency—aligns perfectly with the broader goals of sustainable transportation. As cities worldwide push for greener mobility solutions, technologies that extend battery life and reduce charging frequency will become increasingly valuable. Furthermore, enhanced stability means fewer accidents, lower insurance costs, and greater public confidence in autonomous driving systems, which often rely on similar control architectures.
The implications of this research extend beyond passenger cars. The same principles could be applied to electric buses, delivery vans, and even off-road vehicles, where traction and stability are critical. For instance, in urban transit applications, smoother cornering reduces passenger discomfort and wear on infrastructure. In logistics fleets, improved energy efficiency translates directly into lower operating costs and reduced carbon footprint.
Looking ahead, the next frontier may involve integrating this control strategy with advanced driver assistance systems (ADAS) and vehicle-to-everything (V2X) communication. Imagine a future where your car receives real-time road condition updates from smart infrastructure and adjusts its torque distribution preemptively before entering a slippery curve. Or consider platooning scenarios where multiple electric vehicles coordinate their handling characteristics to maintain tight formation at high speeds.
While the current study focuses on simulation-based validation, the logical next step is hardware-in-the-loop testing and real-world trials. The computational load of the GA-BP neural network and quadratic programming solver must be evaluated against the processing capabilities of modern automotive ECUs. Fortunately, with the rapid advancement of embedded AI chips and real-time operating systems, deploying such algorithms in production vehicles is becoming increasingly feasible.
It’s also worth noting that this control strategy does not require additional sensors beyond those already present in most modern vehicles. It relies on standard inputs such as steering angle, vehicle speed, yaw rate, and lateral acceleration—all of which are routinely measured by electronic stability control (ESC) systems. This compatibility with existing hardware makes the technology highly scalable and cost-effective.
In summary, the work by Zhao and Shi represents a meaningful advancement in the field of electric vehicle dynamics. By rethinking the fundamentals of electronic differential control and embracing modern computational methods, they have created a system that is not only safer and more efficient but also more aligned with the natural behavior of skilled human drivers. As the automotive industry continues its transition to electrification and automation, innovations like this will play a crucial role in shaping the future of mobility.
Their findings underscore a broader trend: the fusion of classical mechanical engineering with cutting-edge information technology is unlocking new possibilities in vehicle performance. It’s no longer enough to build faster or more powerful cars; the challenge now is to make them smarter, more responsive, and more in tune with the environment and the driver.
For engineers, policymakers, and consumers alike, this research offers a glimpse into a future where electric vehicles don’t just replace internal combustion engines—they surpass them in every dimension of performance, safety, and sustainability.
Zhao Nannan, Shi Bo. Torque Electronic Differential Control of Distributed Electric Vehicle. Mechanical Science and Technology for Aerospace Engineering, 2024. DOI: 10.13433/j.cnki.1003-8728.20230085