An active suspension control method based on multi-source perception and passenger comfort optimization

Abstract

The present application belongs to the technical field of vehicle intelligent control, and particularly relates to an active suspension control method based on multi-source sensing and passenger comfort optimization, comprising: S1, obtaining and processing road surface point cloud data and road surface images in front of the vehicle to obtain a comprehensive road surface feature matrix; S2, predicting driving trajectories of left and right front wheels within a future preview time length to generate a road surface elevation profile sequence and a friction coefficient sequence; S3, performing frequency spectrum analysis to obtain road surface excitation frequency spectra corresponding to the left and right front wheels; generating dynamic damping force target curves of adjustable damping shock absorbers of the left and right front wheels through a model predictive control algorithm; S4, performing time compensation to obtain a compensated target damping force instruction sequence; and S5, regulating and controlling the adjustable damping shock absorbers of the left and right front wheels. The method can realize an active suspension response of pre-empting, being accurate, and taking into account passenger comfort and vehicle stability for complex road surface excitation in front of the vehicle during high-speed driving of the vehicle.

Description

Technical Field

[0001] This invention belongs to the field of vehicle intelligent control technology, and in particular relates to an active suspension control method based on multi-source perception and occupant comfort optimization. Background Technology

[0002] In traditional automotive suspension systems, the damping and stiffness parameters of passive suspensions are fixed at the factory and cannot be dynamically adjusted according to driving conditions. This makes it difficult to balance occupant comfort and vehicle handling stability when dealing with complex road surface excitations. To overcome this limitation, semi-active and active suspension systems have been developed. Their core idea is to use sensors such as vehicle acceleration, roll angle, and vertical displacement to provide real-time feedback on the vehicle's dynamic response, thereby adjusting the damping force of the shock absorbers. However, such systems are essentially reactive control—the adjustment action is triggered only after the wheels contact the uneven road surface and the impact has been transmitted to the sprung mass. This results in a significant time lag, especially under high-speed driving or high-frequency, low-amplitude road surface excitations. The control response often lags behind the excitation cycle, weakening the damping effect and potentially exacerbating vibrations due to phase mismatch.

[0003] In recent years, some studies have attempted to introduce forward-looking perception methods, such as millimeter-wave radar and stereo vision cameras, to achieve predictive control. For example, image processing can be used to identify the contours of road shoulders, speed bumps, or potholes, or road surface height information can be used to predict suspension travel. Other solutions map road condition classification results to preset suspension operating modes, such as switching between comfort and sport modes. However, these solutions generally suffer from the following drawbacks:

[0004] (1) Coarse perception granularity: Most rely only on macroscopic road condition classification or discrete event detection, such as speed bump recognition, lacking fine modeling of continuous changes in road surface elevation and local friction characteristics, making it difficult to support the generation of high-precision dynamic damping commands.

[0005] (2) Control logic disconnection: The connection between the sensing module and the suspension execution module is mostly open-loop linkage or threshold triggering, and a closed-loop mapping relationship from road surface micro-terrain → tire-road contact mechanics → suspension dynamic load → occupant comfort index has not been established;

[0006] (3) Uncoupled human factors objectives: Existing predictive strategies usually take minimizing vehicle acceleration or suspension travel constraints as optimization objectives, ignoring the synergistic trade-off between occupant subjective comfort (such as the sensitive area of ​​head acceleration spectrum and dynamic response of seating posture) and vehicle dynamic performance (such as tire contact and lateral stability), resulting in suboptimal control behaviors such as overly soft or stiff response.

[0007] (4) Lack of time delay compensation: Even if the road surface information ahead is obtained, if the inherent delay of the system is not accurately modeled and time axis aligned and compensated, the prediction command will still fail due to phase shift, especially in the typical aiming time domain of T=0.1~0.5s, where the error accumulation is significant.

[0008] Therefore, how to achieve an advanced, precise, and balanced active suspension response to complex road surface excitations while the vehicle is traveling at high speed, without relying on high-cost LiDAR or offline road surface databases, has become an urgent problem to be solved. Summary of the Invention

[0009] To address the aforementioned shortcomings of existing technologies, the present invention aims to provide an active suspension control method based on multi-source perception and occupant comfort optimization. This method enables an active suspension response that is advanced, precise, and balances occupant comfort and vehicle stability in response to complex road surface excitations during high-speed vehicle operation, without relying on high-cost lidar or offline road surface databases.

[0010] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0011] An active suspension control method based on multi-source perception and occupant comfort optimization includes the following steps:

[0012] S1. By installing radar and cameras at the front of the vehicle, real-time point cloud data and road surface images of the road in front of the vehicle are acquired and processed to obtain data including location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix ;

[0013] S2. Combining the current vehicle speed v and steering wheel angle θ, predict the driving trajectories of the left and right front wheels within the future anticipation time T, and obtain the spatial coordinate sequence of the ground contact points corresponding to the left and right front wheels. and Based on the comprehensive pavement feature matrix of S1 The corresponding elevation and friction coefficient are queried respectively to generate a road surface elevation profile sequence. , and friction coefficient sequence , ; t∈[0,T];

[0014] S3, relative to S2 , Spectral analysis was performed to obtain the road excitation spectra corresponding to the left and right front wheels; based on the road excitation spectra and the friction coefficient sequence of S2... , Based on a pre-defined vehicle dynamics model and occupant comfort optimization objectives, a model predictive control algorithm is used to generate dynamic damping force target curves for the adjustable damping shock absorbers of the left and right front wheels. and ;

[0015] S4. Using the system execution delay Δt, for S3 and Time compensation is performed to obtain the compensated target damping force command sequence. and ,in, ;

[0016] S5, obtained using S4 and Adjust the adjustable damping shock absorbers of the left and right front wheels.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. Achieve physical-level feedforward modeling of road surface excitation, overcoming the inherent lag of traditional "reactive" control. Existing semi-active / active suspensions rely on feedback from vehicle body sensors, responding only after the impact has been transmitted to the sprung mass, resulting in control phase lag. This solution fuses forward-looking radar and cameras to construct a comprehensive road surface feature matrix including position, elevation, and friction coefficient. By combining vehicle kinematics to predict the wheel center trajectory, the system can know the excitation characteristics in advance before the wheel contacts the road, thus changing the control from passive suppression to proactive compensation, significantly improving the ability to suppress high-frequency small-amplitude excitations (such as seams and micro-ripples).

[0019] 2. Establish a dynamic coupling mechanism between road surface spectral characteristics and occupant comfort objectives to avoid the limitations of optimizing a single performance index. Existing predictive control often aims to minimize vehicle acceleration or constrain suspension travel, which may sacrifice comfort or handling stability. This solution uses spectral analysis of road surface elevation profile sequences to extract excitation components related to human-sensitive frequency bands, and then models these components together with preset occupant comfort optimization objectives. This embeds human-factor constraints into the damping command generation process, achieving coordinated optimization of comfort, tire contact patch, and lateral stability, rather than a simple compromise.

[0020] 3. A time-axis compensation mechanism ensures the timing effectiveness of predicted commands and resolves the phase mismatch problem caused by execution delays. Even if road condition information is obtained, predicted commands will still fail due to time shift if the total system delay is not corrected. This solution explicitly introduces a delay compensation step (S4) to perform time-shift correction on the damping target curve generated by the model prediction, ensuring that the final output command is strictly synchronized with the actual contact point of the front wheels, making advanced control physically feasible and effectively avoiding control oscillations or gain failure caused by accumulated time delays.

[0021] In summary, this method can achieve an advanced, accurate, and balanced active suspension response to complex road surface excitations while the vehicle is traveling at high speed, without relying on high-cost LiDAR or offline road surface databases.

[0022] Preferably, in step S3, for and Perform Fast Fourier Transform (FFT) on each road surface to obtain the corresponding road surface excitation spectrum. and where f is the frequency;

[0023] When generating the target curve for dynamic damping force, the following strategy is employed:

[0024] 1) Targeting high-frequency excitation components This generates a dynamic damping force command that matches the high-frequency excitation frequency but is opposite in phase, in order to attenuate the unsprung mass vibration caused by high-frequency road surface unevenness; among which, High-frequency threshold;

[0025] 2) Targeting low-frequency excitation components If the coefficient of friction at the corresponding time or Higher than or equal to the preset threshold If the damping force is relatively increased, then a relatively larger damping force is planned to suppress the vertical bounce of the vehicle body; if it is lower than Therefore, a relatively reduced damping force is planned to improve tire contact performance; among which, This is the low-frequency threshold.

[0026] This setup achieves two key advantages: 1. It ensures a physical match between the excitation frequency band and the control strategy, avoiding performance trade-offs caused by a one-size-fits-all approach to damping. Existing active suspensions often employ a uniform damping law (such as PID based on vehicle acceleration feedback), making it difficult to distinguish the frequency domain characteristics of road excitation: high-frequency unevenness can easily trigger unsprung mass resonance, requiring rapid energy dissipation; low-frequency undulations mainly affect vehicle attitude, requiring a balance between buffering and grounding. This solution extracts the excitation spectrum through Fourier transform and sets... , As a demarcation, the high-frequency components trigger anti-phase high damping to attenuate the vibration of the mass block, while the low-frequency components dynamically adjust the damping amplitude based on the real-time friction coefficient—increasing damping to suppress vehicle body bounce when friction is sufficient, and decreasing damping to ensure grounding when friction is insufficient. This strategy decouples the control objectives of different frequency bands from a physical mechanism perspective, which is significantly better than the traditional single-response mode.

[0027] 2. By embedding the tire-road friction coefficient as a decision variable in low-frequency control, the adaptability and safety under complex working conditions are improved. Existing predictive control systems mostly rely solely on geometric road surface information, ignoring changes in contact mechanics; while this solution introduces the real-time friction coefficient in the low-frequency range. , As a threshold criterion, the damping adjustment responds not only to the bumpiness of the road but also to the tire's grip. For example, on wet, slippery surfaces where friction is reduced, the system actively reduces damping to prolong tire contact time and prevent slippage; on dry, high-adhesion surfaces, it increases damping to suppress vehicle sway. This incorporates key contact links in the human-vehicle-road closed loop into the feedforward control chain, significantly improving the system's robustness and safety under unstructured or variable-adhesion road conditions.

[0028] Preferably, in step S3, the objective function J of the model predictive control algorithm is:

[0029] ;

[0030] Where N is the prediction time domain length;

[0031] Let k be the predicted vertical acceleration of the vehicle body at step k;

[0032] The tire dynamic load predicted at step k;

[0033] Let k be the dominant excitation frequency corresponding to the k-th prediction step, derived from... or The frequency band with the largest proportion of medium energy has been determined;

[0034] Weighting coefficient and Dynamically set according to the following rules:

[0035] ;

[0036] ;

[0037] in, , These are the high-frequency comfort penalty weighting coefficient and the low-frequency comfort penalty weighting coefficient, respectively. , These are the high-frequency tire contact penetration penalty weights and the low-frequency tire contact penetration penalty weights, respectively. This indicates that the comfort penalty is weakened at high frequencies; This indicates enhanced tire contact patch control at high frequencies;

[0038] The stronger the penalty for the rate of change of tire dynamic load; This represents the upper limit of the typical friction coefficient for dry asphalt pavement. This is the basic coefficient for grounding penalty.

[0039] This setup achieves a balance between the physical interpretability and adaptability of the target weights, overcoming the shortcomings of traditional fixed-weight MPC in handling complex road conditions. Existing MPC methods typically use fixed weight coefficients, making it difficult to consider different road excitation characteristics: on high-frequency gravel roads, they may excessively suppress vehicle vibration at the expense of ground contact; on low-frequency long-wave undulating roads, they may weaken dynamic load constraints, leading to tire bounce. This solution sets the weights to the excitation frequency. The piecewise function distinguishes between high and low frequencies, and the friction coefficient μ is introduced to modulate the grounding penalty strength in the low-frequency band. The decrease in μ allows the objective function to automatically perceive the current excitation type and adhesion conditions—strengthening comfort penalties and weakening ground contact constraints at high frequencies to avoid tire bounce caused by hard damping; strengthening ground contact penalties to suppress bounce at low frequencies and high adhesion; and actively relaxing ground contact requirements to maintain contact at low frequencies and low adhesion. This mechanism ensures that the optimization objective always aligns with the physical nature and driving safety boundaries, significantly improving the robustness and human factor adaptability of predictive control on non-stable, variable-adhesion road surfaces.

[0040] The objective function design transforms the original empirical and static MPC cost structure into a dynamic decision core with mechanistic support and environmental perception capabilities through a three-level linkage of frequency domain identification, weight switching, and friction modulation. This provides a verifiable and interpretable optimization basis for the intelligent feedforward control of high-order active suspension.

[0041] Preferably, the dominant excitation frequency Determined in the following ways:

[0042] right or In the prediction window Extracting a subsequence from the inner segment and performing a short-time Fourier transform yields the time spectrum. ,in ;

[0043] make frequency band The accumulated energy;

[0044] Dominant incentive frequency Take the satisfaction The lowest frequency, where η is the energy threshold coefficient.

[0045] This setup allows for the physical extraction of the primary excitation frequency, avoiding misjudgments under broadband / multi-peak excitation using the traditional peak frequency method, and ensuring the reliability and adaptability of subsequent frequency division control strategies. Existing predictive control often directly uses the frequency corresponding to the maximum spectral energy as the "primary frequency." However, in real-world road surfaces, such as mixed-type unevenness—containing both long-wave undulations and short-wave gravel—energy peaks may occur in high-frequency noise bands or non-dominant modes. This can lead to the control strategy mistakenly treating high-frequency vibrations as the primary excitation and excessively suppressing them, thus exacerbating the unsprung mass response or sacrificing grounding. This solution introduces an energy threshold constraint η and a minimum frequency priority principle: only when the energy in a certain frequency band reaches a certain proportion of the global maximum value ( Only when the frequency is low are they included in the candidates, and the lowest frequency among them is ultimately selected as the lowest frequency. This design aligns with the fundamental principles of vehicle dynamics—low-frequency components typically dominate overall vehicle movement and occupant comfort perception, while high-frequency components are mostly localized vibrations; through a dual screening process that ensures energy compliance and minimizes frequency, the selected components are optimized for optimal performance. This represents the dominant excitation mode that truly affects the dynamics of the entire vehicle, thus providing an accurate decision-making basis for subsequent frequency-weighted control (such as weakening comfort penalties at high frequencies and strengthening grounding constraints at low frequencies), significantly improving the physical consistency and robustness of the prediction-control link.

[0046] Preferably, when solving the optimization problem of the objective function J, the following damping force change rate constraint is introduced to protect the actuator life:

[0047] ;

[0048] in, γ is the maximum permissible rate of change of damping force of the actuator; γ is the adjustment coefficient; This represents the target damping force output value of the adjustable damping shock absorber of the left / right front wheel during the kth control cycle. The tire-road friction coefficient at the corresponding wheel-end contact point within the current prediction time domain.

[0049] This setup allows for coordinated constraints on the actuator's dynamic response capability and the road surface adhesion state, fundamentally preventing tire bounce or sideslip instability caused by sudden damping changes, and significantly improving control safety under high-risk conditions. Traditional active suspension optimization often only constrains the damping amplitude, neglecting its rate of change—on low-traction surfaces such as ice, snow, and slippery surfaces. If the system suddenly increases the damping (i.e., high dD / dt) to suppress vehicle vibration, it will cause a sudden and drastic reduction in the tire's vertical load, easily inducing tire bounce or even lateral instability; conversely, on high-traction surfaces, the rate of change can be appropriately relaxed to improve response agility. This solution controls the maximum permissible rate of change... and Coupling allows the constraint boundary to adaptively scale with the adhesion conditions: when μ ≪ μ0 (low adhesion), The damping adjustment rate is significantly reduced, preventing aggressive actions from damaging the ground. When μ≈μ0 (high adhesion), the constraint is relaxed, preserving the system's potential for rapid response. This mechanism is not a simple amplitude limiter, but rather a unified model based on the tire-road contact mechanics principle, unifying the actuator's physical limits with the vehicle's safety boundaries. This fundamentally ensures the executability and safety of predictive control commands under extreme conditions.

[0050] Preferably, the high-frequency threshold and low frequency threshold Dynamically adjust based on the current vehicle speed v:

[0051] ;

[0052] ;

[0053] in, This is the preset reference speed; , This is a preset reference threshold;

[0054] When v < 20 km / h, forced =1.0Hz, =12.0 Hz, to cover low-speed, large-fluctuation operating conditions.

[0055] This setup allows for physical alignment of the frequency domain boundary threshold with the vehicle's kinematic characteristics, eliminating control logic mismatch caused by fixed thresholds across speed domains, and ensuring that the frequency division strategy always matches the spatiotemporal scale of the actual excitation. The high and low frequencies of road surface excitation are essentially relative to the vehicle's passing frequency; for example, a pothole with the same wavelength λ=1 m has an excitation frequency of 20 Hz at v=72 km / h (20 m / s), which is high frequency; while at v=18 km / h (5 m / s), it is only 5 Hz, which should be classified as low frequency. If a fixed threshold (such as a constant threshold) is used... =10 Hz), then at high speeds, the true high-frequency vibration will be misjudged as low frequency, weakening the damping suppression; at low speeds, the mid-frequency excitation will be misjudged as high frequency, resulting in excessive suppression, causing a double deterioration in comfort and grounding performance. This solution, based on the kinematic relationship f∝v, makes the threshold scale inversely with vehicle speed, ensuring that at any vehicle speed, and Always corresponds to a physically consistent wavelength range (e.g.) Corresponding to wavelengths >30 m This corresponds to shortwave frequencies <0.5 m, thus ensuring the physical consistency between the subsequent high-frequency anti-phase high-damping and low-frequency friction adaptive damping strategies. Supplemented by a low-speed forced setting, it covers abrupt changes in typical urban operating conditions, further preventing threshold divergence and instability caused by vehicle speed approaching zero. This dynamic threshold mechanism elevates frequency domain division from empirically fixed to kinematically self-consistent.

[0056] Preferably, in step S4, the system execution delay Δt includes the actuator response delay and the computation delay of the sensing-prediction link; or When performing time compensation, an interpolation alignment strategy is adopted, including:

[0057] Let the original instruction sequence be i=0,1,…,M , The sampling interval;

[0058] The compensated instruction at time The value is calculated using linear interpolation:

[0059] ;

[0060] in, To compensate, at time The target damping force command value is output at the location; In the original damping force command sequence, at the sampling time Discrete instruction values; To meet Maximum sampling time; This is the start time of the original instruction sequence.

[0061] This configuration, without increasing hardware response speed, compensates for inherent system latency using pure algorithms, significantly improving the timing accuracy and execution fidelity of control commands and avoiding phase lag and control oscillations caused by delays. In active suspension systems, actuators (such as solenoid valves and MR dampers) have millisecond-level response delays, and predictive-optimization calculations also require several milliseconds. If discrete-time optimization results are directly output under high-speed conditions, the delay can reach 1 / 3 to 1 / 6 of the excitation cycle, causing the control action to lag significantly behind the road excitation, leading to anti-phase suppression failure or even positive feedback oscillations. This solution, through interpolation alignment, enables the controller to achieve synchronization at any given time. The output provides precisely corresponding damping commands—especially between sampling points—by using linear interpolation between adjacent frames of commands. This maintains the physical continuity of the commands (avoiding shocks caused by step jumps) and ensures temporal alignment accuracy. This method requires no additional sensors or advanced prediction models; it relies solely on existing discrete command sequences to significantly improve the effective control bandwidth while suppressing high-frequency resonance peaks in tire dynamic loads, thus significantly improving closed-loop stability and ride quality.

[0062] Preferably, in step S3, the objective of occupant comfort optimization is to minimize the root mean square value of vertical acceleration at the vehicle body center of gravity or the driver's seat rail.

[0063] This setting allows the objective function design to achieve a fundamental leap from minimizing structural response to optimizing human sensory experience, serving as the basis for ensuring the reliability of active suspension comfort performance.

[0064] Preferably, in S1, the processing of the road surface point cloud data and road surface image includes:

[0065] Semantic segmentation is performed on road surface images to identify the road surface material type and condition, and the tire-road friction coefficient μ is estimated accordingly.

[0066] The tire-road friction coefficient μ is spatiotemporally aligned with the road point cloud data to generate a data set containing location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix .

[0067] This setup enables centimeter-level spatial coupling modeling of key tire-road interaction parameters (μ and Z), allowing subsequent predictive control to perform feedforward compensation based on a physically consistent road surface-adhesion joint field. This significantly improves the prediction accuracy and adaptability of the active suspension on non-uniform road surfaces. Traditional methods often isolate road elevation (for vertical excitation prediction) or rely on a single friction estimate (such as wheel speed slip back estimation), resulting in a spatiotemporal mismatch: for example, icy and snowy roads may have gentle elevations (small Z changes), but extremely low μ; while damaged asphalt roads may have drastic Z changes, but still relatively high μ. If only elevation is used to predict the excitation, the risk of low adhesion will be underestimated; if only average μ is used, local abrupt changes (such as oil stains) cannot be distinguished. This scheme uses image semantic segmentation (such as CNN to identify "wet asphalt," "snow," and "oil") to qualitatively and quantitatively map μ(x,y), and then registers it with the Z(x,y) of laser / visual point clouds in the same coordinate system to generate a joint feature matrix. This allows the controller to know the combined working condition of "abrupt elevation change 5 m ahead + friction coefficient drops sharply to 0.2" hundreds of milliseconds before the vehicle arrives, thus adjusting the damping strategy in advance (such as reducing high-frequency stiffness to ensure grounding and suppressing vertical impact) to avoid tire bounce or sideslip caused by passive response.

[0068] Preferably, in step S1, the road surface point cloud data is used to characterize the geometric features of the road surface, and the road surface image is used to characterize the physical features of the road surface. Attached Figure Description

[0069] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0070] Figure 1 This is a flowchart of the method. Detailed Implementation

[0071] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0072] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0073] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figures, or the orientation or positional relationship commonly used when the product is in use. They are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance. In addition, the terms "horizontal," "vertical," etc., do not indicate that the component is required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0074] Example:

[0075] This invention provides an active suspension control method based on multi-source perception and occupant comfort optimization. The invention is applied to an intelligent vehicle chassis control system, which includes a radar and camera mounted at the front of the vehicle, and adjustable damping shock absorbers mounted on each wheel. The radar and camera are jointly mounted within the front bumper, at a height of 0.35-0.65 m above the ground; the adjustable damping shock absorbers are either electromagnetically controlled (e.g., CDC) or magnetorheological (MR) type.

[0076] like Figure 1 As shown, this method includes the following steps:

[0077] S1. By installing radar and cameras at the front of the vehicle, real-time point cloud data and road surface images of the road in front of the vehicle are acquired and processed to obtain data including location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix .

[0078] The road surface point cloud data is used to characterize the geometric features of the road surface, and the road surface image is used to characterize the physical features of the road surface.

[0079] In practice, the processing of road surface point cloud data and road surface images includes:

[0080] Semantic segmentation is performed on road surface images to identify the road surface material type and condition, and the tire-road friction coefficient μ is estimated accordingly.

[0081] The tire-road friction coefficient μ is spatiotemporally aligned with the road point cloud data to generate a data set containing location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix .

[0082] This approach enables centimeter-level spatial coupling modeling of key tire-road interaction parameters (μ and Z), allowing subsequent predictive control to be based on a physically consistent road surface-adhesion joint field for feedforward compensation. This significantly improves the prediction accuracy and adaptability of active suspension on non-uniform road surfaces. Traditional methods often isolate road elevation (for vertical excitation prediction) or rely on a single friction estimate (such as wheel speed slip back estimation), resulting in a spatiotemporal mismatch: for example, icy and snowy roads may have gentle elevations (small Z changes), but extremely low μ; while damaged asphalt roads may have drastic Z changes, but still relatively high μ. If only elevation is used to predict excitation, the risk of low adhesion will be underestimated; if only average μ is used, local abrupt changes (such as oil stains) cannot be distinguished. This solution uses image semantic segmentation (such as CNN to identify "wet asphalt," "snow," and "oil") to qualitatively and quantitatively map μ(x,y), and then registers it with the Z(x,y) of laser / visual point clouds in the same coordinate system to generate a joint feature matrix. This allows the controller to know the combined working condition of "abrupt elevation change 5 m ahead + friction coefficient drops sharply to 0.2" hundreds of milliseconds before the vehicle arrives, thus adjusting the damping strategy in advance (such as reducing high-frequency stiffness to ensure grounding and suppressing vertical impact) to avoid tire bounce or sideslip caused by passive response.

[0083] To facilitate better understanding, image semantic segmentation is explained as follows.

[0084] In one embodiment, the semantic segmentation uses a lightweight U-Net++ network, with a 640×480 RGB image as input and four pixel-level labels as output: {dry asphalt, wet asphalt, snow, oil}; each label corresponds to a pre-defined friction coefficient range.

[0085] Dry bitumen: μ∈ [0.7, 0.9];

[0086] Wet bitumen: μ∈ [0.4, 0.6];

[0087] Snow cover: μ∈ [0.2, 0.3];

[0088] Oil stains: μ∈ [0.1, 0.2];

[0089] For each pixel location (x,y), the center value of its category is taken as the initial value of the local μ(x,y); then, the neighborhood mean filter (window 5×5) is used to smooth the abrupt change to obtain a continuous μ(x,y) field.

[0090] S2. Combining the current vehicle speed v and steering wheel angle θ, predict the driving trajectories of the left and right front wheels within the future anticipation time T, and obtain the spatial coordinate sequence of the ground contact points corresponding to the left and right front wheels. and Based on the comprehensive pavement feature matrix of S1 The corresponding elevation and friction coefficient are queried respectively to generate a road surface elevation profile sequence. , and friction coefficient sequence , ; t∈[0,T].

[0091] S3, relative to S2 , Spectral analysis was performed to obtain the road excitation spectra corresponding to the left and right front wheels; based on the road excitation spectra and the friction coefficient sequence of S2... , Based on a pre-defined vehicle dynamics model and occupant comfort optimization objectives, a model predictive control algorithm is used to generate dynamic damping force target curves for the adjustable damping shock absorbers of the left and right front wheels. and .

[0092] The objective of occupant comfort optimization is to minimize the root mean square value of the vertical acceleration at the vehicle's center of gravity or the driver's seat rail. This objective function design represents a fundamental leap from minimizing structural response to optimizing human experience, and is the basis for ensuring the reliable implementation of active suspension comfort performance.

[0093] In specific implementation, and Perform Fast Fourier Transform (FFT) on each road surface to obtain the corresponding road surface excitation spectrum. and where f is the frequency;

[0094] When generating the target curve for dynamic damping force, the following strategy is employed:

[0095] 1) Targeting high-frequency excitation components This generates a dynamic damping force command that matches the high-frequency excitation frequency but is opposite in phase, in order to attenuate the unsprung mass vibration caused by high-frequency road surface unevenness; among which, This is a high-frequency threshold used to identify high-frequency excitation components.

[0096] 2) Targeting low-frequency excitation components If the coefficient of friction at the corresponding time or Higher than or equal to the preset threshold If the damping force is relatively increased, then a relatively larger damping force is planned to suppress the vertical bounce of the vehicle body; if it is lower than Therefore, a relatively reduced damping force is planned to improve tire contact performance; among which, This is a low-frequency threshold used to identify low-frequency excitation components.

[0097] Existing active suspension systems often employ a uniform damping law (such as PID based on vehicle acceleration feedback), which makes it difficult to distinguish the frequency domain characteristics of road excitation: high-frequency unevenness can easily induce unsprung mass resonance, requiring rapid energy dissipation; low-frequency undulations mainly affect vehicle attitude, requiring a balance between cushioning and grounding. This solution extracts the excitation spectrum through Fourier transform and sets... , As a demarcation, the high-frequency components trigger anti-phase high damping to attenuate the mass block vibration, while the low-frequency components dynamically adjust the damping amplitude based on the real-time friction coefficient—increasing damping to suppress vehicle body bounce when friction is sufficient, and decreasing damping to ensure ground contact when friction is insufficient. This strategy decouples the control objectives of different frequency bands from a physical mechanism perspective, significantly outperforming traditional single-response modes. Furthermore, existing predictive control systems often rely solely on geometric road surface information, ignoring changes in contact mechanics; whereas this solution introduces the real-time friction coefficient in the low-frequency band. , As a threshold criterion, the damping adjustment responds not only to the bumpiness of the road but also to the tire's grip. For example, on wet, slippery surfaces where friction is reduced, the system actively reduces damping to prolong tire contact time and prevent slippage; on dry, high-adhesion surfaces, it increases damping to suppress vehicle sway. This incorporates key contact links in the human-vehicle-road closed loop into the feedforward control chain, significantly improving the system's robustness and safety under unstructured or variable-adhesion road conditions.

[0098] The high frequency threshold and low frequency threshold Dynamically adjust based on the current vehicle speed v:

[0099] ;

[0100] ;

[0101] in, The preset reference speed is 60 km / h; in actual implementation, it will be taken as 60 km / h. , The preset reference thresholds are 2.0Hz and 8.0Hz, respectively, in practice.

[0102] When v < 20 km / h, forced =1.0Hz, =12.0 Hz, to cover low-speed, large-fluctuation operating conditions.

[0103] This setup allows for physical alignment of the frequency domain boundary threshold with the vehicle's kinematic characteristics, eliminating control logic mismatch caused by fixed thresholds across speed domains, and ensuring that the frequency division strategy always matches the spatiotemporal scale of the actual excitation. The high and low frequencies of road surface excitation are essentially relative to the vehicle's passing frequency; for example, a pothole with the same wavelength λ=1 m has an excitation frequency of 20 Hz at v=72 km / h (20 m / s), which is high frequency; while at v=18 km / h (5 m / s), it is only 5 Hz, which should be classified as low frequency. If a fixed threshold (such as a constant threshold) is used... =10 Hz), then at high speeds, the true high-frequency vibration will be misjudged as low frequency, weakening the damping suppression; at low speeds, the mid-frequency excitation will be misjudged as high frequency, resulting in excessive suppression, causing a double deterioration in comfort and grounding performance. This solution, based on the kinematic relationship f∝v, makes the threshold scale inversely with vehicle speed, ensuring that at any vehicle speed, and Always corresponds to a physically consistent wavelength range (e.g.) Corresponding to wavelengths >30 m This corresponds to shortwave frequencies <0.5 m, thus ensuring the physical consistency between the subsequent high-frequency anti-phase high-damping and low-frequency friction adaptive damping strategies. Supplemented by a low-speed forced setting, it covers abrupt changes in typical urban operating conditions, further preventing threshold divergence and instability caused by vehicle speed approaching zero. This dynamic threshold mechanism elevates frequency domain division from empirically fixed to kinematically self-consistent.

[0104] In specific implementation, the objective function J of the model predictive control algorithm is:

[0105] ;

[0106] Where N is the prediction time domain length; Let k be the predicted vertical acceleration of the vehicle body at step k; The tire dynamic load predicted at step k; Let k be the dominant excitation frequency corresponding to the k-th prediction step, derived from... or The frequency band with the largest proportion of medium energy has been determined;

[0107] To facilitate better understanding by those skilled in the art, the design basis of the objective function is explained as follows:

[0108] Vertical acceleration of vehicle body Directly related to passenger comfort, therefore adopting As a cost to comfort;

[0109] Tire dynamic load change rate The stability of tire contact with the ground is determined by sudden loads, which can easily cause tire bounce or sideslip. When the amplitude exceeds a certain value, the risk of tire bounce will increase significantly.

[0110] Weighting coefficient and Dynamic allocation based on excitation frequency band: High-frequency bands focus on suppressing acceleration ( Large), low-frequency band focuses on ensuring grounding ( (Large), conforming to the closed-loop perception characteristics of people-vehicle-road.

[0111] In practical implementation, the weighting coefficient and Dynamically set according to the following rules:

[0112] ;

[0113] ;

[0114] in, , These are the high-frequency comfort penalty weighting coefficient and the low-frequency comfort penalty weighting coefficient, respectively. , These are the high-frequency tire contact penetration penalty weights and the low-frequency tire contact penetration penalty weights, respectively. This indicates that the comfort penalty is weakened at high frequencies; This indicates enhanced tire contact patch control at high frequencies;

[0115] The stronger the penalty for the rate of change of tire dynamic load; This represents the upper limit of the typical friction coefficient for dry asphalt pavement; in practical implementation, it can be taken as 0.9 to 1.0. This is the basic coefficient for grounding penalty.

[0116] Existing MPC methods typically use fixed weighting coefficients, making it difficult to consider different road surface excitation characteristics: on high-frequency gravel roads, they tend to excessively suppress vehicle vibration at the expense of ground contact; on low-frequency long-wave undulating roads, they may weaken dynamic load constraints, leading to tire bounce. This solution sets the weights according to the excitation frequency. The piecewise function distinguishes between high and low frequencies, and the friction coefficient μ is introduced to modulate the grounding penalty strength in the low-frequency band. The reduction in μ allows the objective function to automatically perceive the current excitation type and adhesion conditions—strengthening comfort penalties and weakening ground contact constraints at high frequencies to avoid tire bounce caused by hard damping; strengthening ground contact penalties to suppress bounce at low frequencies and high adhesion; and actively relaxing ground contact requirements to maintain contact at low frequencies and low adhesion. This mechanism ensures that the optimization objective always aligns with the physical nature and driving safety boundaries, significantly improving the robustness and human factor adaptability of predictive control on non-stationary, variable-adhesion road surfaces. This objective function design, through a three-level linkage of frequency domain identification, weight switching, and friction modulation, transforms the originally empirical and static MPC cost structure into a dynamic decision core with mechanistic support and environmental perception capabilities, providing a verifiable and interpretable optimization basis for the intelligent feedforward control of high-order active suspensions.

[0117] In specific implementation, the dominant excitation frequency Determined in the following ways:

[0118] right or In the prediction window Extracting a subsequence from the inner segment and performing a short-time Fourier transform yields the time spectrum. ,in ;

[0119] make frequency band The accumulated energy;

[0120] Dominant incentive frequency Take the satisfaction The lowest frequency, where η is the energy threshold coefficient, specifically η∈[0.3,0.6].

[0121] In a preferred embodiment, the spectrum analysis includes:

[0122] The short-time Fourier transform uses a Hanning window with a window length of Tw = 0.5 s, an overlap rate of 50%, and a frequency resolution of Δf = 1.0 Hz.

[0123] The energy threshold coefficient η is set based on statistical data from actual vehicle road tests, with η = 0.45 being the preferred value;

[0124] Dominant incentive frequency The process of determining is as follows:

[0125] (1) Find all that satisfy Frequency points;

[0126] (2) Perform local maxima detection on these points (no higher value within the neighborhood ±2Δf);

[0127] (3) Among all local maxima, select the one with the lowest frequency as... .

[0128] Existing predictive control often directly takes the frequency corresponding to the maximum spectral energy as the "dominant frequency." However, in real-world road surfaces, such as those with mixed unevenness—containing both long-wave undulations and short-wave gravel—energy peaks may occur in the high-frequency noise range or non-dominant modes. This leads to control strategies mistakenly treating high-frequency vibrations as the primary excitation and excessively suppressing them, thereby exacerbating the unsprung mass response or sacrificing grounding. This solution introduces an energy threshold constraint η and a minimum frequency priority principle: only when the energy in a certain frequency band reaches a certain proportion of the global maximum ( Only when the frequency is low are they included in the candidates, and the lowest frequency among them is ultimately selected as the lowest frequency. This design aligns with the fundamental principles of vehicle dynamics—low-frequency components typically dominate overall vehicle movement and occupant comfort perception, while high-frequency components are mostly localized vibrations; through a dual screening process that ensures energy compliance and minimizes frequency, the selected components are optimized for optimal performance. This represents the dominant excitation mode that truly affects the dynamics of the entire vehicle, thus providing an accurate decision-making basis for subsequent frequency-weighted control (such as weakening comfort penalties at high frequencies and strengthening grounding constraints at low frequencies), significantly improving the physical consistency and robustness of the prediction-control link.

[0129] In practical implementation, when solving the optimization problem of the objective function J, the following damping force change rate constraint is introduced to protect the actuator life:

[0130] ;

[0131] in, γ is the maximum permissible rate of change of damping force of the actuator; γ is the adjustment coefficient, and in specific implementation, γ∈[0,0.5]; This represents the target damping force output value of the adjustable damping shock absorber of the left / right front wheel during the kth control cycle. The tire-road friction coefficient at the corresponding wheel-end contact point within the current prediction time domain.

[0132] Traditional active suspension optimization often only constrains the damping amplitude, neglecting its rate of change—on low-traction surfaces such as ice, snow, and slippery surfaces. If the system suddenly increases the damping (i.e., high dD / dt) to suppress vehicle vibration, it will cause a sudden and drastic reduction in the vertical load on the tires, easily inducing tire bounce or even lateral instability. Conversely, on high-traction surfaces, the rate of change can be appropriately relaxed to improve responsiveness. This solution uses the maximum permissible rate of change... and Coupling allows the constraint boundary to adaptively scale with the adhesion conditions: when μ ≪ μ0 (low adhesion), The damping adjustment rate is significantly reduced, preventing aggressive actions from damaging the ground. When μ≈μ0 (high adhesion), the constraint is relaxed, preserving the system's potential for rapid response. This mechanism is not a simple amplitude limiter, but rather a unified model based on the tire-road contact mechanics principle, unifying the actuator's physical limits with the vehicle's safety boundaries. This fundamentally ensures the executability and safety of predictive control commands under extreme conditions.

[0133] S4. Using the system execution delay Δt, for S3 and Time compensation is performed to obtain the compensated target damping force command sequence. and ,in, This is used to advance the execution time of the target damping force to compensate for system delay.

[0134] In specific implementation, the system execution delay Δt includes the actuator response delay and the computation delay of the sensing-prediction link; for or When performing time compensation, an interpolation alignment strategy is adopted, including:

[0135] Let the original instruction sequence be i=0,1,…,M , The sampling interval; in a preferred embodiment, the sampling interval is... This is equal to the control cycle of the adjustable damping shock absorber, with a typical value of 5 ms;

[0136] The compensated instruction at time The value is calculated using linear interpolation:

[0137] ;

[0138] in, To compensate, at time The target damping force command value is output at the location; In the original damping force command sequence, at the sampling time Discrete instruction values; To meet The maximum sampling time, i.e. (Round down); This is the start time of the original instruction sequence.

[0139] This approach compensates for inherent system latency using pure algorithms without increasing hardware response speed, significantly improving the timing accuracy and execution fidelity of control commands and avoiding phase lag and control oscillations caused by delays. In active suspension systems, actuators (such as solenoid valves and MR dampers) exhibit millisecond-level response delays, while predictive-optimization calculations also require several milliseconds. If discrete-time optimization results are directly output under high-speed conditions, the delay can reach 1 / 3 to 1 / 6 of the excitation cycle, causing the control action to lag significantly behind the road excitation, leading to anti-phase suppression failure or even positive feedback oscillations. This solution, through interpolation alignment, enables the controller to achieve synchronization at any given time. The output provides precisely corresponding damping commands—especially between sampling points—by using linear interpolation between adjacent frames of commands. This maintains the physical continuity of the commands (avoiding shocks caused by step jumps) and ensures temporal alignment accuracy. This method requires no additional sensors or advanced prediction models; it relies solely on existing discrete command sequences to significantly improve the effective control bandwidth while suppressing high-frequency resonance peaks in tire dynamic loads, thus significantly improving closed-loop stability and ride quality.

[0140] S5, obtained using S4 and Adjust the adjustable damping shock absorbers of the left and right front wheels.

[0141] Compared with existing technologies, this invention enables physical-level feedforward modeling of road surface excitation, overcoming the inherent lag of traditional reactive control. Existing semi-active / active suspensions rely on feedback from vehicle body sensors, responding only after the impact has been transmitted to the sprung mass, resulting in control phase lag. This solution, through the fusion of forward-looking radar and cameras, constructs a comprehensive road surface feature matrix including position, elevation, and friction coefficient. By combining vehicle kinematics with prediction of wheel center trajectories, the system can anticipate excitation characteristics before the wheel contacts the road surface, thus shifting control from passive suppression to proactive compensation and significantly improving the ability to suppress high-frequency, small-amplitude excitations (such as seams and micro-undulations). Furthermore, existing predictive control systems often aim to minimize vehicle acceleration or constrain suspension travel, easily sacrificing comfort or handling stability. This solution uses spectral analysis of road elevation profile sequences to extract excitation components related to human-sensitive frequency bands and co-models them with preset occupant comfort optimization goals. This embeds human-factor constraints into the damping command generation process, achieving synergistic optimization of comfort, tire contact patch, and lateral stability, rather than a simple compromise. Moreover, a time-axis compensation mechanism ensures the timing effectiveness of predicted commands, resolving phase mismatch issues caused by execution delays. Even if road condition information is obtained, the predicted command will still fail due to time shift if the total system delay is not corrected. This solution explicitly introduces a delay compensation step (S4) to perform time shift correction on the damping target curve generated by the model prediction, ensuring that the final output command is strictly synchronized with the actual contact point of the front wheels, making advance control physically feasible and effectively avoiding control oscillation or gain failure caused by time delay accumulation.

[0142] This method can achieve an advanced, precise, and balanced active suspension response to complex road surface excitations while the vehicle is traveling at high speed, without relying on high-cost LiDAR or offline road surface databases.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. An active suspension control method based on multi-source perception and occupant comfort optimization, characterized in that, Includes the following steps: S1. By installing radar and cameras at the front of the vehicle, real-time point cloud data and road surface images of the road in front of the vehicle are acquired and processed to obtain data including location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix ; S2. Combining the current vehicle speed v and steering wheel angle θ, predict the driving trajectories of the left and right front wheels within the future anticipation time T, and obtain the spatial coordinate sequence of the ground contact points corresponding to the left and right front wheels. and Based on the comprehensive road surface feature matrix of S1 The corresponding elevation and friction coefficient are queried respectively to generate a road surface elevation profile sequence. , and friction coefficient sequence , ; t∈[0,T]; S3, relative to S2 , Spectral analysis was performed to obtain the road excitation spectra corresponding to the left and right front wheels; based on the road excitation spectra and the friction coefficient sequence of S2... , Based on a pre-defined vehicle dynamics model and occupant comfort optimization objectives, a model predictive control algorithm is used to generate dynamic damping force target curves for the adjustable damping shock absorbers of the left and right front wheels. and ; S4. Using the system execution delay Δt, compared to S3... and Time compensation is performed to obtain the compensated target damping force command sequence. and ,in, ; S5, obtained using S4 and Adjust the adjustable damping shock absorbers of the left and right front wheels.

2. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 1, characterized in that, In step S3, for and Perform Fast Fourier Transform (FFT) on each road surface to obtain the corresponding road surface excitation spectrum. and where f is the frequency; When generating the target curve for dynamic damping force, the following strategy is employed: 1) Targeting high-frequency excitation components This generates a dynamic damping force command that matches the high-frequency excitation frequency but is opposite in phase, in order to attenuate the unsprung mass vibration caused by high-frequency road surface unevenness; among which, High-frequency threshold; 2) Targeting low-frequency excitation components If the coefficient of friction at the corresponding time or Higher than or equal to the preset threshold If the damping force is relatively increased, then a relatively larger damping force is planned to suppress the vertical bounce of the vehicle body; if it is lower than Therefore, a relatively reduced damping force is planned to improve tire contact performance; among which, This is the low-frequency threshold.

3. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 2, characterized in that, In step S3, the objective function J of the model predictive control algorithm is: ； Where N is the prediction time domain length; Let k be the predicted vertical acceleration of the vehicle body at step k; The tire dynamic load predicted at step k; Let k be the dominant excitation frequency corresponding to the k-th prediction step, derived from... or The frequency band with the largest proportion of medium energy has been determined; Weighting coefficient and Dynamically set according to the following rules: ； ； in, , These are the high-frequency comfort penalty weighting coefficient and the low-frequency comfort penalty weighting coefficient, respectively. , These are the high-frequency tire contact penetration penalty weights and the low-frequency tire contact penetration penalty weights, respectively. This indicates that the comfort penalty is weakened at high frequencies; This indicates enhanced tire contact patch control at high frequencies; The stronger the penalty for the rate of change of tire dynamic load; This represents the upper limit of the typical friction coefficient for dry asphalt pavement. This is the basic coefficient for grounding penalty.

4. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 3, characterized in that, The dominant excitation frequency Determined in the following ways: right or In the prediction window Extracting a subsequence from the inner segment and performing a short-time Fourier transform yields the time spectrum. ,in ; make frequency band The accumulated energy; Dominant incentive frequency Take the satisfaction The lowest frequency, where η is the energy threshold coefficient.

5. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 3, characterized in that, When solving the optimization problem of the objective function J, the following damping force change rate constraint is introduced to protect the actuator life: ； in, γ is the maximum permissible rate of change of damping force of the actuator; γ is the adjustment coefficient; This represents the target damping force output value of the adjustable damping shock absorber of the left / right front wheel during the kth control cycle. The tire-road friction coefficient at the corresponding wheel-end contact point within the current prediction time domain.

6. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 2, characterized in that, The high frequency threshold and low frequency threshold Dynamically adjust based on the current vehicle speed v: ； ； in, This is the preset reference speed; , This is a preset reference threshold; When v < 20 km / h, forced =1.0Hz, =12.0 Hz, to cover low-speed, large-fluctuation operating conditions.

7. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 1, characterized in that, In step S4, the system execution delay Δt includes the actuator response delay and the computation delay of the sensing-prediction link; for or When performing time compensation, an interpolation alignment strategy is adopted, including: Let the original instruction sequence be i=0,1,…,M , The sampling interval; The compensated instruction at time The value is calculated using linear interpolation: ； in, To compensate, at time The target damping force command value is output at the location; In the original damping force command sequence, at the sampling time Discrete instruction values; To meet Maximum sampling time; This is the start time of the original instruction sequence.

8. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 1, characterized in that, In step S3, the objective of occupant comfort optimization is to minimize the root mean square value of vertical acceleration at the vehicle body center of gravity or the driver's seat rail.

9. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 1, characterized in that, In S1, the processing of road point cloud data and road images includes: Semantic segmentation is performed on road surface images to identify the road surface material type and condition, and the tire-road friction coefficient μ is estimated accordingly. The tire-road friction coefficient μ is spatiotemporally aligned with the road point cloud data to generate a data set containing location coordinates. Elevation and tire-road friction coefficient Comprehensive pavement feature matrix .

10. The active suspension control method based on multi-source perception and occupant comfort optimization as described in claim 1, characterized in that, In step S1, the road surface point cloud data is used to characterize the geometric features of the road surface, and the road surface image is used to characterize the physical features of the road surface.

Images