Vehicle track smooth reconstruction system based on model and data combination

Abstract

The invention relates to the field of automatic driving data processing, in particular to a vehicle track smooth reconstruction system based on model and data combination, which comprises a data preprocessing module I, a noise reduction module, a data preprocessing module II, a track reconstruction module, a data post-processing module and a training module. According to the invention, a kinematic model reconstruction unit is embedded into a trajectory reconstruction module, the trajectory reconstruction module is trained through data, trajectory reconstruction and driving action extraction are realized, and a noise reduction module is designed to enhance the robustness of the whole system. The system can strictly meet the physical constraint of the vehicle model, the designed network can meet the hard constraint condition, the driving action is learned and extracted, and the track reconstruction and driving action extraction precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving data processing, and particularly to a vehicle trajectory smoothing and reconstruction system based on the combination of a model and data. Background Art

[0002] In fields such as autonomous driving, it is often necessary to sense the trajectory information of surrounding vehicles through sensors. However, the noise of sensors is inevitable, especially in the case of pure vision perception or low-precision perception devices. The trajectory data points polluted by noise will affect downstream tasks such as prediction and decision-making, and it is also difficult to obtain driving actions (acceleration, deceleration, and steering) from the trajectory data. In real engineering, this problem is also reflected in the following situations: the trajectory data sets used for downstream task training usually adopt publicly available data sets collected by high-precision devices and optimized offline. When deployed in real vehicles, due to the limitations of computing power and sensor performance, the trajectory quality drops significantly, resulting in a decline in the performance of the model after deployment.

[0003] Therefore, it is very necessary to smooth the trajectory obtained by the sensor.

[0004] Existing trajectory smoothing methods include filtering methods, fitting methods, and learning methods.

[0005] Among them, filtering methods include Kalman filtering, particle filtering (refer to CN 117973167 A), mean filtering, etc. Usually, the solution efficiency is relatively high and it can run online in real time. However, these methods usually analyze a single trajectory point to obtain the optimal position of a single point. Therefore, it can only reduce noise and improve the jitter of the trajectory, but there will still be significant noise in extracting driving actions from the trajectory. Existing methods for directly obtaining driving actions through vehicle model formulas are very sensitive to noise, and trajectories with a small amount of noise may cause the driving actions to be completely submerged in the noise.

[0006] Fitting methods include polynomial fitting, spline curve fitting, Bezier curve fitting, etc. This type of method has a high degree of trajectory smoothing, but it is limited by the prior of the fitting model, and the diversity of the smoothed trajectory will decrease. In addition, when the form of the selected fitting function is inappropriate, it will cause trajectory distortion, resulting in a single and distorted driving action being extracted. The essence of the fitting problem is an optimization problem, and the solution calculation amount is relatively large. It is more difficult to solve when considering constraints, and the real-time performance is poor.

[0007] Learning methods include inverse reinforcement learning (refer to CN 117407474 A), large language models (refer to CN119377666 A), etc. Usually, they have higher flexibility, but existing models completely rely on data, it is difficult to impose hard constraints, and the reliability is difficult to guarantee. At the same time, using overly complex models is not conducive to actual deployment and use.

[0008] In addition, with the development of end-to-end autonomous driving, the need for obtaining the output of driving actions is becoming increasingly urgent. The existing methods for obtaining driving actions such as acceleration, deceleration, and steering mainly collect data by arranging sensors inside the vehicle. This method has high costs and low efficiency in obtaining driving actions, and the diversity of driving actions is not as rich as the driving data on public roads. Summary of the Invention

[0009] In view of the problems existing in the prior art, the present invention proposes a vehicle trajectory smoothing and reconstruction system based on the combination of a model and data. By introducing a trajectory reconstruction module, the trajectory data is first processed into driving actions, and then a smooth reconstructed trajectory is obtained through kinematic model reconstruction, thereby realizing the reconstruction of the trajectory and the extraction of driving actions. In addition, a noise reduction module is designed to enhance the robustness of the entire system.

[0010] Technical solution of the present invention: A vehicle trajectory smoothing and reconstruction system based on the combination of a model and data, comprising: a first data preprocessing module, a noise reduction module, a second data preprocessing module, a trajectory reconstruction module, a data postprocessing module, and a training module.

[0011] The system inputs a noisy trajectory and outputs a reconstructed trajectory. The reconstructed trajectory is smoother than the input trajectory and strictly satisfies the physical constraints of the kinematic model.

[0012] Specifically, the noisy trajectory first passes through the first data preprocessing module to obtain noisy trajectory features. The first data preprocessing module rotates and translates the trajectory to reduce the trajectory distribution space and improve the learning efficiency.

[0013] The noisy trajectory features pass through the noise reduction module to output a low-noise trajectory. The noise reduction module can reduce noise and ensure the robustness of the subsequent module processing.

[0014] The second data preprocessing module processes the low-noise trajectory and outputs low-noise trajectory features and boundary conditions.

[0015] The low-noise trajectory and boundary conditions are input into the trajectory reconstruction module for processing to obtain a standard reconstructed trajectory.

[0016] The data postprocessing module restores the standard reconstruction to its original position through translation and rotation.

[0017] The training module plays a role during training and is used to train the noise reduction module, the trajectory reconstruction module, and the overall model fine-tuning training.

[0018] Further, the trajectory reconstruction module includes a driving action encoding unit and a kinematic model reconstruction unit. The denoised trajectory features are processed by the driving action encoding unit to obtain driving actions. The driving actions and boundary conditions are input into the kinematic model reconstruction unit to obtain a standard reconstructed trajectory. The driving action encoding unit is a learnable network. After being trained, the driving action encoding unit can process the low-noise trajectory features into driving actions. The kinematic model reconstruction unit is a group of networks built with deterministic operators, which can calculate the standard reconstructed trajectory from the driving actions and boundary conditions and can also propagate gradients during training.

[0019] Further, the vehicle trajectory smoothing reconstruction system based on model and data combination further includes an offline kinematic model reconstruction unit.

[0020] The input and output of the offline kinematic model reconstruction unit are the same as those of the kinematic model reconstruction unit. It can also calculate the standard reconstructed trajectory according to the driving actions and boundary conditions, and has higher accuracy. It can replace the kinematic model reconstruction unit to realize the function of calculating the standard reconstructed trajectory.

[0021] Beneficial effects The present invention proposes a vehicle trajectory smoothing reconstruction system, which can reconstruct a noisy trajectory to obtain a smooth reconstructed trajectory. Compared with traditional methods, the present invention has the following advantages: 1. Considering the physical constraints of the vehicle model: The kinematic model reconstruction unit integrates physical equations into the model training process. Through the design of activation functions, the network can meet the hard constraint conditions.

[0022] 2. The model has strong robustness: By using a preposed noise reduction module, the robustness of the model is enhanced; by adding noise, missing point masks, etc., it can handle situations such as trajectory noise, trajectory frame loss, and incomplete trajectories.

[0023] 3. Improving the accuracy of trajectory reconstruction and driving action extraction: Adopting a bidirectional recursive trajectory reconstruction method enables the gradient of the trajectory error to be propagated bidirectionally during training, improving the accuracy of trajectory reconstruction and driving action extraction. In addition, the offline kinematic model reconstruction unit adopts a two-point reconstruction method to eliminate the sensitive dependence of the initial value recursive reconstructed trajectory on speed and heading, and can obtain a more accurate and robust trajectory reconstruction result. Description of the drawings

[0024] Figure 1 Schematic diagram of the system framework of the present invention; Figure 2 Schematic diagram of the driving action encoding unit of the system of the present invention; Figure 3 Schematic diagram of the definitions of 5 operators in the splitting of the discrete model in the embodiment of the present invention; Figure 4One-way recursive computational graph of the embodiment of the present invention; Figure 5 Two-way recursive computational graph of the embodiment of the present invention; Figure 6 Schematic diagram of the training unit of the noise reduction module of the system of the present invention; Figure 7 Schematic diagram of the training unit of the trajectory reconstruction module of the system of the present invention; Figure 8 Schematic diagram of the overall fine-tuning training unit of the system of the present invention; Figure 9 Schematic diagram of the training strategy of the system of the present invention; Figure 10 Schematic diagram of the trajectory smoothing effect of the embodiment of the present invention ((a) straight trajectory (b) turning trajectory); Figure 11 Schematic diagram of the driving action extraction effect of the embodiment of the present invention ((a) reconstructed trajectory diagram (b) reconstructed acceleration and deceleration actions (c) reconstructed steering actions). Detailed implementation manners

[0025] The technical solution provided by the present application will be further described below in conjunction with specific embodiments and their accompanying drawings. In combination with the following description, the advantages and features of the present application will become clearer.

[0026] As Figure 1 is the overall architecture diagram of the system of the present invention, and the implementation method will be described in detail according to this architecture diagram.

[0027] A vehicle trajectory smoothing and reconstruction system based on the combination of a model and data includes a first data preprocessing module, a noise reduction module, a second data preprocessing module, a trajectory reconstruction module, an offline kinematic model reconstruction unit, a data postprocessing module, and a training module ( Figure 1 not shown in the figure).

[0028] The trajectory reconstruction module includes a kinematic model reconstruction unit and a driving action encoding unit.

[0029] The entire system inputs a noisy trajectory and outputs a reconstructed trajectory. The reconstructed trajectory is smoother than the input trajectory and strictly satisfies the physical constraints of the kinematic model.

[0030] The noisy trajectory first passes through the first data preprocessing module to obtain the noisy trajectory features. The main function of the first data preprocessing module is to rotate and translate the trajectory, reduce the trajectory distribution space, and improve the learning efficiency.

[0031] The noisy trajectory features then pass through the noise reduction module to output a low-noise trajectory. The noise reduction module can reduce noise and ensure the robustness of the subsequent module processing, but the output low-noise trajectory does not strictly satisfy the physical constraints.

[0032] The second data pre - processing module processes the low - noise trajectory and outputs low - noise trajectory features and boundary conditions.

[0033] Both the low - noise trajectory and the boundary conditions enter the trajectory reconstruction module for processing to obtain a standard reconstructed trajectory.

[0034] The trajectory reconstruction module includes a driving action encoding unit and a kinematic model reconstruction unit. The noise - reduced trajectory features are processed by the driving action encoding unit to obtain driving actions. The driving actions and boundary conditions are input into the kinematic model reconstruction unit to obtain a standard reconstructed trajectory.

[0035] The driving action encoding unit is a learnable network. After being trained, the driving action encoding unit can process low - noise trajectory features into driving actions.

[0036] The kinematic model reconstruction unit is a set of computational networks built with deterministic operators. It calculates the standard reconstructed trajectory according to the driving actions and boundary conditions and can propagate gradients during training.

[0037] The data post - processing module restores the standard reconstruction to its original position through translation and rotation.

[0038] The input and output of the offline kinematic model reconstruction unit are the same as those of the kinematic model reconstruction unit. It can also calculate the standard reconstructed trajectory according to the driving actions and boundary conditions and has higher accuracy. After the trajectory reconstruction module is trained, the kinematic model reconstruction unit can be replaced with the offline kinematic model reconstruction unit to obtain higher accuracy.

[0039] The training module plays a role during training and needs to call the above - mentioned first data pre - processing module, noise reduction module, second data pre - processing module, and trajectory reconstruction module.

[0040] The following details each functional module, where: The first data pre - processing module: First, the noise trajectory data needs to be pre - processed. The trajectory is translated and rotated to move the end point of the trajectory to the origin and rotate the trajectory so that the heading angle of the end point is 0. After processing, noise trajectory features are obtained. The purpose of this data pre - processing is to compress the distribution space of the trajectory and improve the learning efficiency (prior art); The noise reduction module: After being trained, it can output a low - noise trajectory when inputting noise trajectory features. The noise reduction module processing includes: first, flattening the trajectory tensor, then processing it through a multi - layer perceptron layer (mlp), and finally restoring the shape of the output tensor (unflatten).

[0041] Among them, is the input noise trajectory feature, To output a low-noise trajectory.

[0042] The second data preprocessing module processes the low-noise trajectory output by the noise reduction module. Since driving actions are different from trajectory points and there is no spatial position dependence, the coordinates of adjacent trajectory points of the low-noise trajectory are differentially calculated to obtain the low-noise trajectory features and compress the data distribution. In addition, boundary conditions need to be calculated for use by the kinematic model reconstruction unit or the offline kinematic model reconstruction unit. The boundary conditions include the coordinates (x, y) of the trajectory boundary points and the states (v, φ) of the boundary points.

[0043] The boundary points can be the starting point, the ending point or any point on the trajectory. The number of boundary points and the selection of boundary points vary according to the requirements of the reconstruction algorithm. In the following embodiments, the initial value one-way recursive trajectory reconstruction only requires the coordinates and states of one boundary point, while the initial value two-way recursive trajectory reconstruction method requires the coordinates of two boundary points and the state of one of the boundary points.

[0044] The method for calculating the state of a certain boundary point through discrete coordinates is as follows In the formula and are the boundary point and the adjacent point of the boundary point, is the speed, is the heading angle, is the trajectory point sampling period.

[0045] The function of the trajectory reconstruction module is to process the trajectory and obtain a usable vehicle kinematic model to analyze the smooth trajectory of the driving action. Limited by the vehicle system and real physical conditions, the trajectory of an actual vehicle cannot have severe jitters and jumps. The function of the trajectory reconstruction module is to first obtain a driving action that meets the physical conditions and then reconstruct the driving action into a vehicle trajectory. Completing these two operations can obtain a trajectory that can be smoothly analyzed by the vehicle kinematic model. These two operations are respectively completed by the driving action encoding unit and the kinematic model reconstruction unit.

[0046] The trained driving action encoding unit can process the low-noise trajectory features into driving actions.

[0047] The driving action encoding unit is composed of a multi-layer perceptron and an activation function layer, as Figure 2 shown: The driving action encoding unit is a learnable network Among them, F is the functional form representation of the driving action encoding unit, represents the learnable parameters, is the driving action, Including acceleration and deceleration actions and steering actions , Denote constraint parameters: , wherein, is the upper bound of the vehicle's acceleration and deceleration actions, is the upper bound of the steering action.

[0048] Specifically, a multi-layer perceptron (MLP) can be selected to obtain the intermediate features of the driving actions (including the intermediate features of acceleration and deceleration and the intermediate features of steering ). wherein, is the low-noise trajectory feature processed by the second data preprocessing module.

[0049] The intermediate features are encoded by a multi-layer perceptron and the driving actions that meet the hard constraint limitations are obtained through an activation function with upper and lower limits. The activation function layer needs to have symmetric upper and lower limits, and optionally, the hyperbolic tangent activation function is adopted to ensure that the output meets the hard constraints The constraint boundaries of acceleration and deceleration can be selected , and the constraint boundaries of steering .

[0050] The kinematic model reconstruction unit is designed in the following way: Theoretical part (deduction) First, start the derivation from the vehicle kinematic model: The vehicle kinematic model with the center of the rear axle of the vehicle as the reference point can be expressed as wherein, the ratio of the tangent of the front wheel angle and the wheelbase is defined as the equivalent steering input . It can be proved through theoretical derivation that the absolute value of the equivalent steering input is the same as the curvature of the trajectory.

[0051] Then discretize: Since the trajectory points are discrete, a discrete model needs to be used. Therefore, after discretizing the above model, we get: wherein, and are the horizontal and vertical coordinates of the vehicle at the k-th and (k + 1)-th steps; and are the speeds of the vehicle at the k-th and (k + 1)-th steps; and are the heading angles of the vehicle at the k-th and (k + 1)-th steps; is the discrete step length, which is consistent with the sampling period in the dataset.

[0052] Next, the discrete model is split, and five operators are designed, including: the forward trajectory coordinate recursion operator, the forward state recursion operator, the backward trajectory coordinate recursion operator, the backward state recursion operator, and the averaging operator, where: is the forward trajectory coordinate recursion operator, is the forward state recursion operator, is the backward trajectory coordinate recursion operator, is the backward state recursion operator, represents the averaging operator, The above forward means that during trajectory reconstruction, the recursion starts from the trajectory coordinate point with the smallest time, and the backward means that the recursion starts from the trajectory coordinate point with the largest time.

[0053] The expressions of each operator are as Figure 3 shown.

[0054] The function that the kinematic model reconstruction unit needs to complete is to construct the trajectory back from the driving action where, is the driving action, is the boundary condition, is the reconstructed trajectory, is the kinematic trajectory recursion, which is a deterministic process without learnable parameters in the middle.

[0055] Specifically, the kinematic model reconstruction unit uses the above operators, combines the driving action and the boundary condition, and reconstructs the trajectory. There are many ways for the kinematic model reconstruction unit to reconstruct the trajectory, and different recursive reconstruction methods have different effects during error backpropagation and thus different effects during training.

[0056] As an example, the kinematic model reconstruction unit is implemented using the initial value single-direction recursive trajectory reconstruction and the initial value two-direction recursive trajectory reconstruction respectively.

[0057] For the initial value single-direction recursive trajectory reconstruction, the computational graph is as Figure 4 shown. The recursion can start from the end point or the start point, and the processing process is similar. Figure 4The computational graph shown is a computational graph for forward recursive calculation starting from the end point: First, the boundary conditions at the end point include the boundary point coordinates and the boundary point state . and are input into the reverse trajectory coordinate recursion operator g1 to obtain the coordinate .

[0058] The boundary point state and the driving action are input into the reverse state recursion operator g2 to obtain the state .

[0059] Similar recursive calculations are performed on the driving action ; until the calculation is performed until the end point position, and the complete trajectory coordinates ; … , and the states ; … are obtained.

[0060] Through the above recursive calculations, the complete trajectory coordinate points, that is, the standard reconstructed trajectory, are obtained.

[0061] The above-mentioned initial value bidirectional recursive trajectory reconstruction has a computational graph as Figure 5 shown.

[0062] Similar to the initial value forward recursive reconstruction, it can start from the end point or start from the end point. The processing process is similar and the effect is approximate, which is selected according to the requirements. As Figure 5 shown, taking the recursion starting from the end point as an example: The first half of the recursive process is the same as the initial value forward recursive trajectory reconstruction, and the complete trajectory coordinates ; … , and the states ; … are obtained.

[0063] Subsequently, calculations are performed again with and as the boundary conditions. Among them, uses the boundary conditions processed by the second data preprocessing module, while is obtained from the results of the forward segment-by-segment recursive process.

[0064] The purpose of using the results obtained from the step-by-step recursive process from the front is to ensure the consistency of the trajectory shape. For the initial value two-way recursive trajectory reconstruction method, the state of one end is used as the input in the other direction. In this way, a boundary point state and two boundary point coordinates are required as the initial value conditions. The forward and reverse reconstructed trajectory shapes are the same, and the calculated average trajectory shapes are also the same.

[0065] As Figure 5 shown, and are input into the forward trajectory coordinate recursive operator f1, and is calculated. The boundary point state and the driving action are input into the forward state recursive operator f2 to obtain the state .

[0066] Similar recursive calculations are performed on the driving action ; Until the calculation is performed until the starting position, and the complete trajectory coordinates can be obtained; …, .

[0067] Finally, the complete trajectories obtained in the two directions are averaged to obtain the final output trajectory, that is, the standard reconstructed trajectory.

[0068] The characteristics of the above two methods are as follows: For the initial value one-way recursive trajectory reconstruction method, the trajectory point error can only be propagated backward to the driving actions in front of it, while for the initial value two-way recursive trajectory reconstruction method, the error gradient can be propagated from two directions.

[0069] The training module includes: a noise reduction module training unit, a trajectory reconstruction module training unit, and an overall fine-tuning training unit, which are used to train the noise reduction module, the trajectory reconstruction module, and the overall fine-tuning training of the model, respectively. Specifically as follows: As Figure 6 shown, the input of the noise reduction module training unit is the trajectory of the trajectory dataset, and the output is the pre-training parameters of the noise reduction module. The noise reduction module training unit first performs data contamination processing on the dataset trajectory: find a high-quality trajectory dataset and slice the trajectory into the same appropriate length. Subsequently, the trajectory is contaminated, and the processing means include adding various forms of random noise, randomly discarding some data, and incomplete segment data. Record the masks for the positions of the discarded and incomplete data.

[0070] The contaminated noisy trajectory is used as the input of the noise reduction module, and the original high-quality trajectory data is used as the label. The lost frames and incomplete data are calculated using the masked labels.

[0071] The noise reduction module training subsystem includes: a data pre - processing module 1, a noise reduction module, and a data post - processing module. After being trained, the noise reduction module training unit can obtain the pre - trained parameters of the noise reduction module.

[0072] The loss function of the noise reduction module training unit is where N is the trajectory length, that is, the number of coordinate points in the trajectory, is the trajectory output by the noise reduction module, is the standard trajectory obtained after passing through the data pre - processing module 1, that is, the input trajectory of the noise reduction module.

[0073] As Figure 7 shown, the input of the trajectory reconstruction module training unit is the dataset trajectory, and the output is the pre - trained parameters of the driving action encoding unit. The trajectory reconstruction module training unit inputs the dataset trajectory into the trajectory reconstruction module training subsystem. The dataset trajectory is used as a label for training. The trajectory reconstruction module training subsystem includes: a data pre - processing module 1, a data pre - processing module 2, a trajectory reconstruction module, and a data post - processing module. After being trained, the trajectory reconstruction module training unit can obtain the pre - trained parameters of the driving action encoding unit.

[0074] When training the trajectory reconstruction module, two factors are considered. One is to define the MSE error of trajectory reconstruction based on the proximity of the reconstructed trajectory where is the trajectory output by the trajectory reconstruction module, is the standard trajectory obtained after passing through the data pre - processing module 1.

[0075] The other is for the smoothness of driving actions, which is measured by the average action amount. Usually, when completing the same task, human drivers tend to use the minimum action amount rather than frequent high - frequency adjustments. Therefore, the loss of driving action smoothness is designed as where is the acceleration and deceleration action loss, is the steering action loss.

[0076] The final training loss function is composed of these two parts where and are weight coefficients, and the selectable range is , in this embodiment, is selected, .

[0077] As Figure 8 shown, the input of the overall fine-tuning training unit is the trajectory data set trajectory, the pre-trained parameters of the noise reduction module, and the pre-trained parameters of the driving action encoding unit, and the output is the final parameters of the noise reduction module and the final parameters of the driving action encoding unit. The overall fine-tuning training unit processes the data set trajectory for data contamination and then inputs it into the overall fine-tuning training subsystem. The data set trajectory is used as a label for training. The overall fine-tuning training subsystem includes: a data pre-processing module 1, a noise reduction module, a data pre-processing module 2, a trajectory reconstruction module, and a data post-processing module. The overall fine-tuning training unit uses the pre-trained parameters of the noise reduction module and the pre-trained parameters of the driving action encoding unit as initial parameters, and can obtain the final parameters of the noise reduction module and the final parameters of the driving action encoding unit after fine-tuning training.

[0078] The loss function during the training of the overall fine-tuning training unit is the same as that of the trajectory reconstruction module training unit.

[0079] As Figure 9 shown, the training strategy is to first train the noise reduction module training unit and the trajectory reconstruction module training unit. Then, the pre-trained parameters obtained from the training are passed to the overall fine-tuning training unit for training.

[0080] In addition, since the boundary point state (v, φ) needs to be calculated indirectly through coordinates, and the initial boundary point state (v, φ) has a greater impact on the trajectory reconstruction result and is prone to amplifying errors during the recursive process. Therefore, an offline kinematic model reconstruction unit is designed, which has higher computational accuracy (but a larger amount of computation). After the model training is completed, it can replace the kinematic model reconstruction unit to complete the trajectory reconstruction (or the kinematic model reconstruction unit can also be continued to be used).

[0081] The offline kinematic model reconstruction unit adopts a two-point reconstruction method, that is, it reconstructs according to the driving action output by the driving action encoding unit and the boundary point coordinates of the starting point and the ending point. The boundary point coordinates are more accurate than the boundary point state, and the trajectory point coordinate noise has a smaller impact on the reconstruction result. However, the two-point reconstruction method cannot directly obtain the coordinates of each intermediate point. Therefore, first, the boundary point state is solved using the boundary point coordinates of the starting point and the ending point, and then the complete reconstructed trajectory is obtained by recursion. Solving the boundary point state requires solving the following non-linear equation In the formula, is the starting point trajectory coordinate, is the ending point trajectory coordinate.

[0082] Perform variable substitution , Thus, separating variables gives Subsequently, it is calculated in two cases When The non - linear equation to be solved is And Or When, the equation is It is calculated in two cases and obtained Then solve It is difficult to obtain an analytical solution for the above - mentioned non - linear equation, and numerical methods can be used for solution. Therefore, it can only be reconstructed in an offline manner.

[0083] After obtaining the initial state 、 After that, the initial value can be used for one - way recursive trajectory reconstruction.

[0084] The trajectory results extracted in this embodiment are as Figure 10 shown. Among them, in the figure, (a) is the straight - line trajectory, and (b) is the turning trajectory. The results show that the method proposed by the present invention can effectively reduce noise and obtain a smooth trajectory. The blue curve in the figure is the high - quality trajectory in the dataset, and the green curve is the noisy trajectory after contaminating the original data. The red curve is the reconstructed trajectory processed by the vehicle trajectory smoothing and reconstruction system described in the present invention.

[0085] The driving action results extracted in this embodiment are as Figure 11 shown. Among them, in the figure, (a) is the trajectory before and after reconstruction, (b) is the acceleration and deceleration action, and (c) is the steering action. Figure 11 The blue curve in it is the high - quality trajectory and the driving action calculated directly by the differential method. The red curve is the reconstructed trajectory and the driving action output by the driving action coding unit in the trajectory reconstruction module, and the black dotted line is the physical constraint.

[0086] From Figure 10 And Figure 11 The implementation effects, it can be concluded that this method can effectively smooth the jittery vehicle trajectory, and the smoothed trajectory can strictly meet the vehicle physical constraints. This processing method can effectively extract driving actions.

[0087] The above description is only a description of the preferred embodiments of the present application, and is not any limitation on the scope of the present application. Any change or modification made by any person skilled in the art according to the disclosed technical content should be regarded as an equivalent effective embodiment, and all belong to the scope protected by the technical solution of the present application.

[0088] Appendix: Glossary:[[]] ​Driving actions: Actions for operating a vehicle to move, including steering, accelerating, braking, etc. Multilayer Perceptron (MLP): A feedforward neural network that contains one or more hidden layers and is capable of learning non-linear models to perform classification or regression analysis on input data.

Claims

1. A vehicle trajectory smoothing reconstruction system based on model and data combination, characterized in that: include: Data pre-processing module 1, noise reduction module, data pre-processing module 2, trajectory reconstruction module, data post-processing module and training module; the system inputs the noisy trajectory and outputs the reconstructed trajectory. The reconstructed trajectory is smoother than the input trajectory and strictly meets the physical constraints of the kinematic model; Specifically, The noise trajectory first passes through the data pre-processing module to obtain the noise trajectory feature; the data pre-processing module rotates and translates the trajectory to reduce the trajectory distribution space and improve learning efficiency; The noise trajectory features pass through the noise reduction module to output a low-noise trajectory; the noise reduction module can reduce noise and ensure the robustness of post-module processing; The second data pre-processing module processes the low-noise trajectory and outputs the low-noise trajectory characteristics and boundary conditions; The low-noise trajectory and boundary conditions are input into the trajectory reconstruction module for processing to obtain a standard reconstructed trajectory; The data post-processing module restores the standard reconstruction to the original position through translation and rotation; The training module plays a role during training and is used to train the denoising module, the trajectory reconstruction module and the overall fine-tuning training of the model.

2. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 1, characterized in that: The trajectory reconstruction module includes a driving action encoding unit and a kinematic model reconstruction unit; the denoised trajectory feature is passed through the driving action encoding unit to obtain the driving action, and the driving action and boundary conditions are input into the kinematic model reconstruction unit to obtain a standard reconstructed trajectory; The driving action encoding unit is a learnable network. After training, the driving action encoding unit can process low-noise trajectory features into driving actions; the kinematic model reconstruction unit is a group of networks built with certain operators, which can calculate standard reconstructed trajectories based on driving actions and boundary conditions.

3. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 1, characterized in that: The denoising module processes as follows: first flatten the trajectory tensor, then process it with the multi-layer perceptron layer mlp, and finally restore the shape of the output tensor to unflatten: in, is the input noise trajectory feature, To output a low noise trajectory.

4. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 1, characterized in that: The data pre-processing module II processes as follows: The coordinates of adjacent track points of low-noise track are calculated differentially to obtain low-noise track features and compress data distribution; Calculate boundary conditions and provide them to a kinematic model reconstruction unit or an offline kinematic model reconstruction unit, wherein the boundary conditions include trajectory boundary point coordinates (x, y) and boundary point states (v, φ); The method for calculating the state of a boundary point through discrete coordinates is as follows In the formula and are the boundary points and their adjacent points, For speed, is the heading angle, is the trajectory point sampling period.

5. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 2, characterized in that: The driving action encoding unit is composed of a multi-layer perceptron and an activation function layer; The driving action encoding unit is a learnable network as follows in, F is the functional form of the driving action encoding unit, represents the learnable parameters, For driving action, Including acceleration and deceleration and steering action , Represents constraint parameters: ， in, is the upper limit of vehicle acceleration and deceleration action, is the upper bound of the turning action; Select multi-layer perceptron mlp to obtain the intermediate features of driving actions, including acceleration and deceleration intermediate features and turn to intermediate features : in, The low-noise trajectory features obtained by processing the second data pre-processing module; The intermediate features are encoded by a multi-layer perceptron, and the driving action that satisfies the hard constraints is obtained through an activation function with upper and lower limits. The activation function layer uses a hyperbolic tangent activation function. To ensure that the output meets the hard constraints Select the acceleration and deceleration constraint boundaries , the constraint boundary of the steering .

6. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 2, characterized in that: The kinematic model reconstruction unit is designed in the following manner: Theoretical part: First, we start with the vehicle kinematic model: The vehicle kinematic model with the center of the vehicle rear axle as the reference point can be expressed as The ratio of the front wheel turning angle tangent to the wheelbase is defined as is the equivalent steering input ;Through theoretical derivation, it can be proved that the absolute value of the equivalent steering input is the same as the curvature of the trajectory; Then discretize: Because the trajectory points are discrete, a discrete model is needed, so the above model is discretized to obtain: in, and are the horizontal and vertical coordinates of the vehicle at the kth and k+1th steps; and is the speed of the vehicle at the kth and k+1th steps; and is the heading angle of the vehicle at the kth and k+1th steps; is the discrete step length, which is consistent with the sampling period in the data set; Next, the discrete model was split and five operators were designed, including: forward trajectory coordinate recursion operator, forward state recursion operator, reverse trajectory coordinate recursion operator, reverse state recursion operator and averaging operator; The above forward direction means that the trajectory reconstruction starts from the trajectory coordinate point with the smallest time, and the reverse direction means that the reconstruction starts from the trajectory coordinate point with the largest time. The kinematic model reconstruction unit completes the function of constructing the trajectory from the driving action: in, For driving action, is the boundary condition, To reconstruct the trajectory, It is a recursive kinematic trajectory, which is a deterministic process with no learnable parameters. The kinematic model reconstruction unit reconstructs the trajectory by using the above operator in combination with the driving action and the boundary conditions.

7. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 6, characterized in that: The kinematic model reconstruction unit reconstructs the trajectory in two ways: initial value one-way recursive trajectory reconstruction and initial value two-way recursive trajectory reconstruction. Different recursive reconstruction methods have different effects during error back propagation and thus different effects during training.

8. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 1, characterized in that: The training module includes: a denoising module training unit, a trajectory reconstruction module training unit and an overall fine-tuning training unit, which are respectively used to train the denoising module, the trajectory reconstruction module and the overall fine-tuning training of the model; as follows: The input of the denoising module training unit is the trajectory of the trajectory data set, and the output is the pre-training parameters of the denoising module; the denoising module training unit processes the trajectory of the dataset for data contamination and then inputs it into the denoising module training subsystem; the trajectory of the dataset is used as a label for training; the denoising module training subsystem includes: a data pre-processing module, a denoising module and a data post-processing module; the denoising module training unit can obtain the pre-training parameters of the denoising module after training; The loss function of the denoising module training unit is in, N is the trajectory length, i.e. the number of coordinate points in the trajectory, is the trajectory output by the denoising module, is the standard trajectory obtained after the data pre-processing module, that is, the input trajectory of the noise reduction module; The trajectory reconstruction module training unit inputs the data set trajectory and outputs the driving action encoding unit pre-training parameters. The trajectory reconstruction module training unit inputs the data set trajectory into the trajectory reconstruction module training subsystem; the data set trajectory is used as a label for training; the trajectory reconstruction module training subsystem includes: a data pre-processing module 1, a data pre-processing module 2, a trajectory reconstruction module and a data post-processing module; the trajectory reconstruction module training unit can obtain the driving action encoding unit pre-training parameters after training; Two factors are considered when training the trajectory reconstruction module. One is the MSE error of the trajectory reconstruction defined by the degree of closeness of the reconstructed trajectory. in, is the trajectory output by the trajectory reconstruction module, It is the standard trajectory obtained after the data pre-processing module; Second, the smoothness of driving action is measured by the average amount of action. Usually, when completing the same task, human drivers tend to use the smallest amount of action rather than high-frequency repeated adjustments. Therefore, the loss of driving action smoothness is designed to be in is the acceleration and deceleration action loss, Loss of steering action; The final training loss function consists of these two parts in, and is the weight coefficient; The overall fine-tuning training unit inputs the trajectory of the trajectory data set, the pre-training parameters of the denoising module, and the pre-training parameters of the driving action encoding unit, and outputs the final parameters of the denoising module and the final parameters of the driving action encoding unit; the overall fine-tuning training unit performs data pollution processing on the trajectory of the data set and inputs it into the overall fine-tuning training subsystem; the trajectory of the data set is used as a label for training; the overall fine-tuning training subsystem includes: a data pre-processing module, a denoising module, a data pre-processing module, a trajectory reconstruction module, and a data post-processing module; the overall fine-tuning training unit uses the pre-training parameters of the denoising module and the pre-training parameters of the driving action encoding unit as initial parameters, and can obtain the final parameters of the denoising module and the final parameters of the driving action encoding unit after fine-tuning training; The loss function during training of the overall fine-tuning training unit is the same as that of the trajectory reconstruction module training unit; The training strategy is to first train the denoising module training unit and the trajectory reconstruction module training unit, and then pass the trained pre-training parameters to the overall fine-tuning training unit for training.

9. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 2, characterized in that: It also includes an offline kinematic model reconstruction unit; The input and output of the offline kinematic model reconstruction unit are consistent with those of the kinematic model reconstruction unit, and a standard reconstructed trajectory is calculated according to driving actions and boundary conditions with higher accuracy, replacing the kinematic model reconstruction unit to realize the function of calculating the standard reconstructed trajectory.

10. A vehicle trajectory smoothing reconstruction system based on model and data combination as claimed in claim 9, characterized in that: The offline kinematic model reconstruction unit adopts a two-point reconstruction method, that is, reconstruction is performed according to the driving action output by the driving action encoding unit and the coordinates of the boundary points of the starting point and the end point; First, the boundary point state is solved using the coordinates of the starting and ending boundary points, and then the complete reconstructed trajectory is obtained recursively. Solving the boundary point state requires solving the following nonlinear equation: In the formula, is the starting point trajectory coordinate, is the coordinate of the end point trajectory; Perform variable substitution , Thus, the variables are separated and obtained Then the calculation is divided into two cases: when The nonlinear equation to be solved is: and or When , the equation is: Calculate in two cases Post-Solve The above nonlinear equations are solved by numerical methods; Solve to get the initial state , Afterwards, the initial value one-way recursive trajectory is used to reconstruct.

Images