New Control Strategy Enhances Ride Comfort in Hub Motor EVs
In the rapidly evolving landscape of electric mobility, engineers and researchers are continuously seeking innovative solutions to overcome the inherent challenges associated with advanced drivetrain technologies. Among these, hub motor-driven electric vehicles (EVs) have emerged as a promising design due to their compact architecture, high energy efficiency, and dynamic controllability. However, the integration of motors directly into the wheel assembly introduces new engineering complexities—most notably, increased unsprung mass and the generation of unbalanced electromagnetic forces under operational eccentricities. These factors can significantly degrade ride comfort and vehicle stability, posing a critical barrier to widespread adoption.
A recent breakthrough from Jiangsu University’s School of Automotive and Traffic Engineering offers a compelling answer to this challenge. Professor Zhongxing Li, along with his graduate student Xue Wang and colleagues Qiqian Cheng and Yi Yu, has developed a novel damping control strategy for air suspension systems in hub motor EVs. Their work, published in the Journal of Chongqing University of Technology (Natural Science), introduces a hybrid control framework that combines Linear Quadratic Regulator (LQR) theory with Genetic Algorithm (GA) optimization—termed GA-LQR control—to dynamically manage suspension damping in response to complex vibrational inputs.
The research addresses a core issue in modern EV design: the trade-off between performance innovation and ride quality. While hub motors eliminate traditional drivetrain components such as gearboxes and differentials, thereby reducing mechanical losses and freeing up interior space, they also place additional mass at the extremities of the vehicle’s suspension system. This increase in non-sprung mass reduces the suspension’s ability to isolate the cabin from road disturbances. Moreover, when manufacturing tolerances, road irregularities, or load variations cause misalignment between the motor’s stator and rotor, an uneven air gap is created. This imbalance distorts the magnetic field, generating oscillating electromagnetic forces that act directly on the wheel hub, further exciting the suspension and amplifying vertical vibrations.
Traditional passive suspension systems are ill-equipped to handle such dynamic disturbances. Even conventional semi-active systems, which adjust damping in real time based on sensor feedback, often rely on heuristic or rule-based control logic that lacks global optimality. In contrast, optimal control strategies like LQR offer a mathematically rigorous approach to balancing multiple performance objectives—such as minimizing body acceleration, controlling tire load variation, and limiting suspension travel—within a unified framework.
However, as the authors point out, the effectiveness of LQR control hinges critically on the selection of weighting matrices—Q, which governs the relative importance of each state variable (e.g., body acceleration, pitch rate, suspension deflection), and R, which penalizes control effort. Historically, these parameters have been tuned through trial and error or based on engineering intuition, making the process subjective and suboptimal. Recognizing this limitation, the team from Jiangsu University turned to evolutionary computation to automate and enhance the tuning process.
Genetic Algorithms (GAs), inspired by the principles of natural selection and genetic inheritance, are particularly well-suited for navigating high-dimensional, non-linear optimization landscapes. By encoding potential solutions as “chromosomes,” applying crossover and mutation operators, and iteratively selecting the fittest individuals across generations, GAs can efficiently converge on near-optimal parameter sets without requiring gradient information or assuming convexity.
In this study, the researchers implemented a real-coded GA to search for the ideal combination of Q and R matrix elements within predefined bounds. The fitness function was carefully designed to reflect real-world performance metrics: the root mean square (RMS) values of sprung mass vertical acceleration, hub motor eccentricity, suspension working space, tire dynamic load, and pitch angular acceleration—all normalized against baseline passive suspension performance. This normalization ensured that the optimization process remained robust across varying driving conditions and avoided bias due to differences in measurement units.
To validate the theoretical model before implementing the control strategy, the team constructed an 8-degree-of-freedom half-vehicle dynamic model that explicitly accounts for the structural and electromagnetic interactions between the hub motor and the air suspension system. Unlike simplified quarter-car models, this half-car configuration captures both vertical bounce and pitch motions, offering a more realistic representation of vehicle behavior. The model incorporates detailed subcomponents: the stator mass, rotor and rim assembly, tire ring mass, air spring force, variable damping, and crucially, the electromagnetic excitation force arising from rotor eccentricity.
The accuracy of this simulation framework was verified through real-world testing. A production electric vehicle was retrofitted with integrated air suspension units featuring magnetorheological dampers—fluid-based shock absorbers whose viscosity can be altered rapidly via an applied magnetic field, enabling precise, real-time control of damping force. Accelerometers were mounted at strategic locations—on both the top and bottom mounts of the rear air spring—to capture the vertical acceleration of the sprung and unsprung masses. Data was collected during controlled runs over a B-class random road surface at a constant speed of 30 km/h, using a professional LMS data acquisition system.
Frequency-domain analysis of the test results showed a strong correlation between the simulated and measured acceleration profiles. While minor discrepancies existed—attributed to unmodeled factors such as rubber bushing compliance and assembly tolerances—the overall trend and peak amplitudes aligned closely. This empirical validation gave the researchers confidence in using the model for subsequent control development and performance evaluation.
With the model confirmed, the team proceeded to analyze the impact of unbalanced electromagnetic forces on vehicle dynamics. Simulations were conducted under B-class road conditions at speeds ranging from 50 to 80 km/h, comparing scenarios with and without electromagnetic excitation. The results were striking: at 80 km/h, the RMS body acceleration increased by 26.63% due to motor eccentricity, rising from 0.73 m/s² to 0.99 m/s². Similarly, tire dynamic load surged by 21.53%, and pitch acceleration climbed by 24.45%. These findings underscore the non-negligible influence of electromagnetic disturbances—often overlooked in conventional vehicle dynamics studies—and justify the need for advanced control interventions.
The GA-LQR controller was then implemented in a MATLAB/Simulink environment, with the optimization performed offline to ensure real-time feasibility during vehicle operation. This approach allowed the generation of a pre-calibrated lookup table mapping specific driving conditions—such as road class and vehicle speed—to optimal weighting parameters. During actual driving, the controller could rapidly retrieve the appropriate Q and R values, apply the precomputed feedback gain matrix, and command the magnetorheological dampers to deliver the ideal damping force.
Performance comparisons were conducted under two key scenarios: varying road roughness (B-class vs. C-class) at a fixed speed of 60 km/h, and varying speed (50–80 km/h) on a B-class road. In all cases, the GA-LQR strategy outperformed both passive suspension and standard LQR control. On a B-class road at 60 km/h, the GA-LQR system reduced body acceleration by 23.83%, motor eccentricity by 22.90%, tire load by 16.36%, and pitch acceleration by 9.65%—all relative to the passive baseline. In contrast, conventional LQR achieved improvements of only 10.11%, 11.63%, 8.68%, and 6.54% respectively.
Even more impressive was the controller’s robustness under deteriorating conditions. On a rougher C-class road, where vibration energy is significantly higher, the GA-LQR maintained strong performance, reducing body acceleration by 20.61% compared to passive suspension, while standard LQR managed only 7.73%. This resilience highlights the advantage of GA-based tuning: by exploring a broader solution space, the algorithm discovers parameter combinations that provide better trade-offs under extreme conditions, whereas manually tuned LQR controllers often prioritize nominal performance at the expense of robustness.
Time-domain simulations across multiple speeds further confirmed the superiority of the GA-LQR approach. As vehicle speed increased, so did the severity of vibration inputs, leading to higher RMS values across all metrics. Yet, the GA-LQR controller consistently delivered greater attenuation than its LQR counterpart, demonstrating its adaptability and effectiveness across a range of operating points.
To assess the impact on human comfort, the researchers conducted a frequency-domain analysis of the simulation results, focusing on the 4–8 Hz range identified by ISO 2631-1 as the most sensitive to vertical vibration. In this critical band, the GA-LQR control reduced the power spectral density (PSD) of body acceleration from a peak of 0.13 (m/s²)²/Hz under passive suspension to 0.09 (m/s²)²/Hz—a 30.8% reduction. Similar improvements were observed in other metrics: motor eccentricity PSD dropped from 3.28×10⁻⁵ to 2.70×10⁻⁵ m²/Hz, tire load PSD from 150 to 94 N²/Hz, and pitch acceleration PSD to 0.098 (rad/s²)²/Hz in the 0–5 Hz range.
These spectral reductions translate directly into tangible benefits for passengers. Lower vibration levels in the 4–8 Hz range reduce fatigue, improve occupant well-being, and enhance the perceived quality of the ride. For luxury EVs and autonomous vehicles—where cabin comfort is a key selling point—such advancements could become a differentiating factor in a competitive market.
Beyond comfort, the GA-LQR strategy also contributes to vehicle safety and component longevity. By reducing tire dynamic load, the control system helps maintain consistent road contact, improving traction and braking performance, especially on uneven surfaces. Lower suspension travel minimizes the risk of bottoming out, while reduced motor eccentricity may extend the lifespan of bearings and reduce electromagnetic noise and losses.
The implications of this research extend beyond hub motor EVs. The GA-LQR framework is inherently scalable and adaptable. It could be applied to other types of active and semi-active suspension systems, including those in conventional vehicles, commercial trucks, or off-road machinery. The methodology of combining physics-based modeling with data-driven optimization represents a paradigm shift in automotive control engineering—one that moves away from heuristic design toward systematic, performance-guaranteed solutions.
Moreover, the use of offline optimization with online lookup tables strikes a practical balance between computational complexity and real-time responsiveness. As vehicle electronic control units (ECUs) become more powerful and connected, such hybrid approaches could be integrated with predictive road preview systems, using camera or LiDAR data to anticipate upcoming road conditions and pre-adjust suspension parameters for optimal performance.
In conclusion, the work by Li, Wang, Cheng, and Yu presents a comprehensive and rigorously validated solution to one of the most pressing challenges in next-generation electric vehicle design. By integrating electromagnetic modeling, multi-body dynamics, optimal control theory, and evolutionary computation, they have developed a control strategy that not only mitigates the drawbacks of hub motor integration but also elevates the overall driving experience. Their findings demonstrate that with intelligent control, the disadvantages of advanced propulsion systems can be effectively managed, paving the way for wider adoption of innovative EV architectures.
As the automotive industry continues its transition toward electrification and automation, research like this will play a crucial role in ensuring that technological progress does not come at the expense of comfort, safety, or reliability. The GA-LQR control method stands as a testament to the power of interdisciplinary engineering—where mechanical, electrical, and control systems converge to create smarter, smoother, and more sustainable transportation solutions.
Zhongxing Li, Xue Wang, Qiqian Cheng, Yi Yu, School of Automotive and Traffic Engineering, Jiangsu University. Journal of Chongqing University of Technology (Natural Science), doi: 10.3969/j.issn.1674-8425(z).2024.03.002