Steering-by-wire road feel adaptive control method based on environmental semantics and driving style

Abstract

The application is suitable for the technical field of automobile chassis control, and provides a steer-by-wire road feel adaptive control method based on environmental semantics and driving style, which comprises the following steps: collecting vehicle external environment information and driving state information; calculating characteristic parameters based on the vehicle driving state information to obtain a steering style coefficient; outputting a virtual vibration torque signal according to the road surface material type; combining the steering style coefficient and the virtual vibration torque signal to calculate a target steering hand force; and finally, closed-loop control of the motor outputting the hand force. Through the fusion of road surface environmental semantics and driving style, the application can generate real texture touch according to the road surface material and dynamically adjust the road feel feedback strength according to the driver's preference, thereby solving the problem of single road feel and poor scene adaptability of the existing steer-by-wire system and significantly improving the realism and personalized experience of driving.

Description

Technical Field

[0001] This invention belongs to the field of automotive chassis control technology, and particularly relates to a steer-by-wire adaptive control method based on environmental semantics and driving style. Background Technology

[0002] With the rapid development of automobiles towards intelligence and electrification, steer-by-wire (SBW) systems, by eliminating the mechanical connection between the steering wheel and the steering actuator, have become one of the core technologies for achieving high-level autonomous driving and intelligent chassis. However, the elimination of the mechanical connection also brings a fundamental problem: the driver can no longer directly perceive the real feedback from the road surface through the steering system, resulting in a so-called loss of "road feel".

[0003] To address this challenge, the industry has proposed several technical solutions. For example, Chinese invention patent CN119117092B, entitled "Road Feel Feedback Control Method, Device, Electronic Equipment, Storage Medium, and Vehicle," obtains rack force through Kalman filtering combined with an equivalent dynamic model, and performs interval and frequency-division filtering based on vehicle speed to obtain road feel feedback torque. After friction compensation and road feel weight adjustment, the road feel motor outputs reaction force. While this method can achieve a certain degree of fluctuation suppression and personalized adjustment, its road feel adaptation relies solely on the single parameter of vehicle speed, failing to consider the driver's driving style and thus struggling to meet the differentiated needs of users with different driving styles.

[0004] For example, the Chinese invention patent CN121448508A, entitled "A Method, System, Medium and Device for Adaptive Design of Road Feel in Steer-by-Wire," pre-trains a road feel neural network by collecting various data from benchmark vehicles and then incrementally trains and adjusts it on a real vehicle by combining the driver's subjective evaluation. Although this method can adapt to driver preferences, its data collection cost is high, the training cycle is lengthy, and it does not fully consider the differences in road feel caused by different road surface types (such as asphalt and icy roads), resulting in insufficient scene adaptability.

[0005] In summary, existing technical solutions have failed to effectively integrate environmental semantic information (such as road surface material and roughness) with the driver's driving style to achieve dynamic and personalized adaptive road feel adjustment. Therefore, there is an urgent need in this field for a steer-by-wire road feel control method that can solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a steer-by-wire adaptive control method based on environmental semantics and driving style, aiming to solve the problems mentioned in the background art.

[0007] The embodiments of the present invention are implemented as follows: a steer-by-wire adaptive control method based on environmental semantics and driving style includes the following steps:

[0008] Step 1: Information collection, synchronously collecting information on the vehicle's external environment and driving status;

[0009] Step 2: Driving style recognition. Based on driving status information, calculate the feature parameters that characterize the driver's handling characteristics to obtain the steering style coefficient;

[0010] Step 3: Virtual road sense construction, identifying the current road surface type based on the vehicle's external environment information, and generating virtual vibration torque signals according to the road surface type;

[0011] Step 4: Target steering force synthesis: Combine the steering style coefficient and the virtual vibration torque signal to calculate the target steering force;

[0012] Step 5: Closed-loop control, using the target steering force as a reference input to drive the motor of the steer-by-wire feel simulator, thereby reproducing the target road feel.

[0013] A further technical solution involves collecting vehicle driving status information, including vehicle speed, in step 1. Steering wheel angle Steering wheel angular velocity Steering wheel operating torque yaw rate Lateral acceleration .

[0014] A further technical solution involves acquiring road image data via an onboard camera in step 1. And acquire road point cloud data through vehicle-mounted LiDAR. :

[0015]

[0016] in, These are point cloud spatial coordinates;

[0017] Using a deep learning-based semantic segmentation network, the current road surface type is identified and output. According to the road surface type Determine the roughness coefficient from the preset environmental semantic transformation table. Predicted adhesion coefficient and environmental gain coefficient ;

[0018] Finally, the environmental semantic parameters are obtained. :

[0019] .

[0020] In a further technical solution, step 2 includes the following specific steps:

[0021] Set time window Within a time window, statistical calculations are performed on vehicle driving status data to extract feature parameters characterizing driver handling characteristics. Including the average steering wheel angular velocity Steering wheel reversal frequency Average steering wheel torque and peak lateral acceleration :

[0022]

[0023]

[0024]

[0025]

[0026] in, It is the number of sampling points within the time window; It is the steering wheel angular velocity at the i-th sampling moment; It is the number of times the sign of the steering wheel angular velocity changes within the time window; It is the steering wheel torque at the i-th sampling time; It is a function that takes the maximum value;

[0027] A driving behavior feature vector is constructed using the aforementioned feature parameters. :

[0028]

[0029] The feature parameters are normalized so that their values ​​are within the interval [0,1]. The normalization formula is as follows:

[0030]

[0031] in, These are the normalized feature parameters (corresponding to) , , and ); These are the minimum and maximum values ​​of the feature parameter;

[0032] The normalized feature parameters are weighted and fused to calculate the style shift coefficient. :

[0033]

[0034] in, These are feature weight coefficients that satisfy... ;

[0035] Steering style coefficient The range of values ​​for is [0,1].

[0036] Time smoothing is applied to the steering style coefficient:

[0037]

[0038] in, It is the steering style coefficient of the current time window after smoothing; It is the steering style coefficient calculated in the current time window; It is the steering style coefficient of the previous time window after smoothing; It is the smoothing coefficient. .

[0039] In step 2, the time window length The length is 15 seconds.

[0040] In a further technical solution, step 3 includes the following specific steps:

[0041] The power spectral density of the road surface was calculated using the ISO 8608 model, and the calculation method is as follows:

[0042]

[0043] in, It is spatial frequency. , For spatial frequency range; It is the reference spatial frequency; It is the roughness coefficient; It is a frequency index;

[0044] spatial frequency range Divided into K intervals, virtual vibration torque signal The calculation is as follows:

[0045]

[0046] in, It is the environmental gain coefficient; These are the vibration amplitudes of each frequency component; It is the real-time frequency of each frequency component; It is a time variable; Is it size in Random numbers between;

[0047]

[0048] in, It is the value at the center frequency of the k-th interval of the power spectral density; It is the frequency range in which the center frequency point is located;

[0049]

[0050] in, It is the center spatial frequency of the k-th interval; This is the current vehicle speed.

[0051] In a further technical solution, step 4 includes the following specific steps:

[0052] Calculate the basic restoring torque based on a simplified version of the magic formula. :

[0053]

[0054] in, It is the lateral force of the front wheel; It is a pneumatic trail. It is mechanical trailing distance;

[0055] The lateral force of the tire is calculated using a simplified linear model:

[0056]

[0057] in, It refers to tire lateral stiffness; It is the front wheel slip angle;

[0058]

[0059] in, It's the front wheel steering angle; It is the distance from the center of gravity to the front axle;

[0060]

[0061] in, It is the steering gear ratio;

[0062] A saturation function is introduced to handle the basic restoring torque:

[0063]

[0064] in, It is a saturation function; It is the restoring torque after saturation treatment; It is the calculated basic restoring torque; It is the preset maximum return torque threshold;

[0065] To compensate for the internal frictional resistance of the steering system, a frictional compensation torque based on the LuGre model is introduced. :

[0066]

[0067]

[0068] in, It is the average deformation of the bristles, which is an internal state variable; These are the stiffness and damping coefficients; For the Stribeck function;

[0069] The steering style coefficient for the current time window, after smoothing, is calculated based on step 2. The restoring torque after saturation treatment Virtual vibration torque signal and friction compensation torque Perform weighted fusion to obtain the target turning hand force. :

[0070]

[0071] in, and It is a regulatory factor; The larger the value, the greater the restoring torque after saturation treatment. The proportion is reduced, and the virtual vibration torque signal The larger.

[0072] In a further technical solution, the current command output to the motor in step 5 is calculated as follows:

[0073]

[0074]

[0075] in, It is the current command output to the motor; It is torque tracking error; It is the calculated current target steering force; It is the steering wheel torque measured in real time by the torque sensor; It is the closed-loop control coefficient.

[0076] The adaptive road feel control method for steer-by-wire based on environmental semantics and driving style provided in this invention has the following beneficial effects:

[0077] (1) Improve the realism of road feel and environmental adaptability: By integrating the environmental semantic information perceived by the vehicle camera and lidar, it can accurately identify different road surface materials such as asphalt, cement, gravel, and snow, and convert their unique texture features (characterized by roughness coefficient, power spectral density, etc.) into virtual vibration torque signals. This makes the simulated road feel more realistic and delicate, and significantly enhances the system's adaptability to different driving scenarios.

[0078] (2) Achieving highly personalized road feel matching: By analyzing the driver's steering operation data in real time (such as angular velocity, reversal frequency, and steering torque), a driving style recognition model is constructed to obtain a quantified steering style coefficient. This coefficient serves as the core adjustment parameter and is dynamically integrated into the synthesis process of the target steering force. As a result, the system can automatically provide clearer and more direct road feel feedback for drivers who prefer an "aggressive" driving style, and provide a lighter and more comfortable feel for drivers who prefer a "smooth" driving style, achieving true "personalized feel for each driver" and significantly improving driving comfort and the human-machine co-driving experience.

[0079] (3) Balancing system safety and computational efficiency: It does not rely on high-cost pre-trained neural networks or a large amount of real vehicle calibration data. Its method based on physical models (such as ISO 8608 and LuGre models) and statistical characteristics has low computational cost and is easy to implement in engineering. At the same time, by saturating the basic return torque and smoothing the steering style coefficient, the stability and safety of the control output are ensured. Attached Figure Description

[0080] Figure 1 A flowchart of the steer-by-wire road feel adaptive control method based on environmental semantics and driving style provided in an embodiment of the present invention;

[0081] Figure 2 A comparison chart of target steering force waveforms under different driving styles. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0083] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0084] like Figure 1As shown, a steer-by-wire adaptive control method based on environmental semantics and driving style, provided in an embodiment of the present invention, includes the following steps:

[0085] Step 1: Information Collection;

[0086] Simultaneously collecting vehicle external environment information and driving status information provides a data foundation for driving style recognition and virtual road sense construction. The specific steps are as follows:

[0087] Collect vehicle driving status information, including vehicle speed. Steering wheel angle Steering wheel angular velocity Steering wheel operating torque yaw rate Lateral acceleration .

[0088] Obtain road image data through vehicle-mounted cameras And acquire road point cloud data through vehicle-mounted LiDAR. :

[0089]

[0090] in, These are the spatial coordinates of the point cloud.

[0091] Using a deep learning-based semantic segmentation network, the current road surface type is identified and output. ( According to road surface type Determine the roughness coefficient from the preset environmental semantic transformation table. Predicted adhesion coefficient and environmental gain coefficient The environmental semantic translation table is shown in Table 1 below:

[0092] Table 1. Environmental Semantic Transformation Table

[0093] Road surface type <![CDATA[Roughness coefficient (10 -6 m 3 )]]> Predicted adhesion coefficient Environmental gain coefficient asphalt 64 (ISO Class B) 0.85 1.0 cement 256 (ISO Class C) 0.80 1.2 gravel 4096 (ISO Class E) 0.60 1.6 Ice and snow 16 (Texture is covered) 0.25 0.4

[0094] Finally, the environmental semantic parameters are obtained. :

[0095]

[0096] Step 2: Driving style recognition;

[0097] Based on the vehicle driving status information collected in step one, characteristic parameters representing the driver's handling characteristics are calculated to obtain the steering style coefficient. The specific steps are as follows:

[0098] Set time window The length is 15 seconds in this invention.

[0099] Within a time window, statistical calculations are performed on vehicle driving status data to extract feature parameters that characterize the driver's operating characteristics. Including the average steering wheel angular velocity Steering wheel reversal frequency Average steering wheel torque and peak lateral acceleration :

[0100]

[0101]

[0102]

[0103]

[0104] in, It is the number of sampling points within the time window; It is the steering wheel angular velocity at the i-th sampling moment; It is the number of times the sign of the steering wheel angular velocity changes within the time window; It is the steering wheel torque at the i-th sampling time; It is a function that takes the maximum value.

[0105] A driving behavior feature vector is constructed using the aforementioned feature parameters. :

[0106]

[0107] Since different feature parameters have different dimensions, each feature parameter is normalized to ensure its value range is within the interval [0,1]. The normalization formula is as follows:

[0108]

[0109] in, These are the normalized feature parameters (corresponding to) , , and ); These are the minimum and maximum values ​​of the feature parameter.

[0110] The normalized feature parameters are weighted and fused to calculate the style shift coefficient. :

[0111]

[0112] in, These are feature weight coefficients that satisfy... .

[0113] Steering style coefficient The value range is [0, 1]. When A smaller value indicates that the driver's handling is relatively smooth, tending towards a conservative style; when... A larger value indicates that the driver's steering is more aggressive, approaching an aggressive style.

[0114] To avoid abrupt changes in driving style recognition results, the steering style coefficient is smoothed over time.

[0115]

[0116] in, It is the steering style coefficient of the current time window after smoothing; It is the steering style coefficient calculated in the current time window; It is the steering style coefficient of the previous time window after smoothing; It is the smoothing coefficient. .

[0117] Step 3: Virtual Road Sense Construction; Based on the road surface material type identified in Step 1, output a virtual vibration torque signal containing a specific frequency and amplitude. The steps are as follows:

[0118] The power spectral density of the road surface was calculated using the ISO 8608 model, and the calculation method is as follows:

[0119]

[0120] in, It is spatial frequency. , For the spatial frequency range, take per meter; It is the reference spatial frequency, taken as per meter; It is the roughness coefficient, which is read from the environment semantic table; It is the frequency exponent, which is set to 2.

[0121] spatial frequency range Divided into K intervals. Virtual vibration torque signal. The calculation is as follows:

[0122]

[0123] in, It is the environmental gain coefficient, which represents the required road feel feedback intensity for different road surfaces. It can be obtained by looking up the environmental semantic table according to different road conditions. These are the vibration amplitudes of each frequency component; It is the real-time frequency of each frequency component; It is a time variable; Is it size in Random numbers between these ranges are used to ensure the randomness of the signal.

[0124]

[0125] in, It is the value at the center frequency of the k-th interval of the power spectral density; It is the frequency range in which the center frequency point is located.

[0126]

[0127] in, It is the center spatial frequency of the k-th interval; This is the current vehicle speed.

[0128] Step 4: Target steering force synthesis. Combine the steering style coefficient obtained in Step 2. The virtual vibration torque signal generated in step 3 Calculate the steering force of the target The specific steps are as follows:

[0129] Calculate the basic restoring torque based on a simplified version of the magic formula. :

[0130]

[0131] in, It is the lateral force of the front wheel; It is the aerodynamic trail, which decreases as the sideslip angle increases; It is the mechanical trailing distance, which is a constant.

[0132] The lateral force of the tire is calculated using a simplified linear model:

[0133]

[0134] in, It refers to tire lateral stiffness; It is the front wheel slip angle.

[0135]

[0136] in, It's the front wheel steering angle; It is the distance from the center of mass to the front axle.

[0137]

[0138] in, It is the steering gear ratio.

[0139] To avoid excessively large foundation restoring torque, a saturation function is introduced to handle it:

[0140]

[0141] in, It is a saturation function; It is the restoring torque after saturation treatment; It is the calculated basic restoring torque; It is the preset maximum return torque threshold, which is set to 10 Newton-meters.

[0142] To compensate for the internal frictional resistance of the steering system, a frictional compensation torque based on the LuGre model is introduced. :

[0143]

[0144]

[0145] in, It is the average deformation of the bristles, which is an internal state variable; These are the stiffness and damping coefficients; This is the Stribeck function, which is related to the steering wheel angular velocity and is used to describe the nonlinear characteristics of friction torque as a function of speed in the low-speed region.

[0146] The steering style coefficient for the current time window, after smoothing, is calculated based on step 2. The restoring torque after saturation treatment Virtual vibration torque signal and friction compensation torque By performing weighted fusion, the target turning force is obtained:

[0147]

[0148] in, and It is a regulatory factor. The larger the value, the greater the restoring torque after saturation treatment. The proportion is reduced, reducing the feeling of heaviness, and the virtual vibration torque signal is used. The larger the size, the clearer the road feel.

[0149] Step 5: Closed-loop control; Apply the calculations obtained in Step 4... As a reference input, the motor of the steer-by-wire simulator is driven by a current loop controller. Closed-loop adjustment is performed based on the deviation between the actual output torque of the motor and the target torque to achieve high-precision reproduction of the target road feel. The steps are as follows:

[0150] The current command output to the motor is calculated as follows:

[0151]

[0152]

[0153] in, It is the current command output to the motor; It is torque tracking error; It is the calculated current target steering force; It is the steering wheel torque measured in real time by the torque sensor; It is the closed-loop control coefficient.

[0154] To verify the effectiveness of the steer-by-wire adaptive road feel control method based on the fusion of environmental semantics and driving style described in this invention, a simulation analysis was conducted using typical daily driving conditions. Figure 2 This is a comparison chart of the target steering force waveforms for different driving styles under the conditions of asphalt road surface, vehicle speed of 60km / h, and continuous small-amplitude sinusoidal steering input. Figure 2 It is evident that, under the same steering input and road conditions, a conservative driving style... The corresponding target steering effort is minimal, resulting in a smooth steering feel; neutral driving style. The corresponding target steering effort amplitude is moderate, balancing steering smoothness and road feel intensity; aggressive driving style The corresponding target steering force amplitude is the largest, and the road feel feedback is clearer and more direct. The results show that this invention, through the steering style coefficient... It can adaptively adjust the target steering force amplitude, so that the road feel intensity can be adaptively matched with the driving style, satisfying the handling preferences and road feel needs of drivers with different styles.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A steer-by-wire adaptive road feel control method based on environmental semantics and driving style, characterized in that, Includes the following steps: Step 1: Information collection, synchronously collecting information on the vehicle's external environment and driving status; Step 2: Driving style recognition. Based on driving status information, calculate the feature parameters that characterize the driver's handling characteristics to obtain the steering style coefficient; Step 3: Virtual road sense construction, identifying the current road surface type based on the vehicle's external environment information, and generating virtual vibration torque signals according to the road surface type; Step 4: Target steering force synthesis: Combine the steering style coefficient and the virtual vibration torque signal to calculate the target steering force; Step 5: Closed-loop control, using the target steering force as a reference input to drive the motor of the steer-by-wire feel simulator, thereby reproducing the target road feel.

2. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 1, characterized in that, In step 1, vehicle driving status information is collected, including vehicle speed. Steering wheel angle Steering wheel angular velocity Steering wheel operating torque yaw rate Lateral acceleration .

3. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 2, characterized in that, In step 1, road image data is acquired through the vehicle-mounted camera. And acquire road point cloud data through vehicle-mounted LiDAR. : in, These are point cloud spatial coordinates; Using a deep learning-based semantic segmentation network, the current road surface type is identified and output. According to the road surface type Determine the roughness coefficient from the preset environmental semantic transformation table. Predicted adhesion coefficient and environmental gain coefficient ; Finally, the environmental semantic parameters are obtained. : 。 4. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 3, characterized in that, Step 2 includes the following specific steps: Set time window Within a time window, statistical calculations are performed on vehicle driving status data to extract feature parameters characterizing driver handling characteristics. Including the average steering wheel angular velocity Steering wheel reversal frequency Average steering wheel torque and peak lateral acceleration : in, It is the number of sampling points within the time window; It is the steering wheel angular velocity at the i-th sampling moment; It is the number of times the sign of the steering wheel angular velocity changes within the time window; It is the steering wheel torque at the i-th sampling time; It is a function that takes the maximum value; A driving behavior feature vector is constructed using the aforementioned feature parameters. : The feature parameters are normalized so that their values ​​are within the interval [0,1]. The normalization formula is as follows: in, These are the normalized feature parameters; These are the minimum and maximum values ​​of the feature parameter; The normalized feature parameters are weighted and fused to calculate the style shift coefficient. : in, These are feature weight coefficients that satisfy... ; Steering style coefficient The range of values ​​for is [0,1]. Time smoothing is applied to the steering style coefficient: in, It is the steering style coefficient of the current time window after smoothing; It is the steering style coefficient calculated in the current time window; It is the steering style coefficient of the previous time window after smoothing; It is the smoothing coefficient. .

5. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 4, characterized in that, Time window length The length is 15 seconds.

6. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 4, characterized in that, Step 3 includes the following specific steps: The power spectral density of the road surface was calculated using the ISO 8608 model, and the calculation method is as follows: in, It is spatial frequency. , For spatial frequency range; It is the reference spatial frequency; It is the roughness coefficient; It is a frequency index; spatial frequency range Divided into K intervals, virtual vibration torque signal The calculation is as follows: in, It is the environmental gain coefficient; These are the vibration amplitudes of each frequency component; It is the real-time frequency of each frequency component; It is a time variable; Is it size in Random numbers between; in, It is the value at the center frequency of the k-th interval of the power spectral density; It is the frequency range in which the center frequency point is located; in, It is the center spatial frequency of the k-th interval; This is the current vehicle speed.

7. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 6, characterized in that, Step 4 includes the following specific steps: Calculate the basic restoring torque based on a simplified version of the magic formula. : in, It is the lateral force of the front wheel; It is a pneumatic trail. It is mechanical trailing distance; The lateral force of the tire is calculated using a simplified linear model: in, It refers to tire lateral stiffness; It is the front wheel slip angle; in, It's the front wheel steering angle; It is the distance from the center of gravity to the front axle; in, It is the steering gear ratio; A saturation function is introduced to handle the basic restoring torque: in, It is a saturation function; It is the restoring torque after saturation treatment; It is the calculated basic restoring torque; It is the preset maximum return torque threshold; To compensate for the internal frictional resistance of the steering system, a frictional compensation torque based on the LuGre model is introduced. : in, It is the average deformation of the bristles, which is an internal state variable; These are the stiffness and damping coefficients; For the Stribeck function; The steering style coefficient for the current time window, after smoothing, is calculated based on step 2. The restoring torque after saturation treatment Virtual vibration torque signal and friction compensation torque Perform weighted fusion to obtain the target turning hand force. : in, and It is a regulatory factor; The larger the value, the greater the restoring torque after saturation treatment. The proportion is reduced, and the virtual vibration torque signal The larger.

8. The steer-by-wire adaptive road feel control method based on environmental semantics and driving style according to claim 7, characterized in that, In step 5, the current command output to the motor is calculated as follows: in, It is the current command output to the motor; It is torque tracking error; It is the calculated current target steering force; It is the steering wheel torque measured in real time by the torque sensor; It is the closed-loop control coefficient.

Images