Vehicle automatic driving method and system, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a vehicle automatic driving method and system, electronic equipment and a storage medium, and belongs to the technical field of vehicle automatic driving. According to the method, electroencephalogram signals of a driver are collected, a braking decision of the driver is recognized in real time according to the electroencephalogram signals, then path planning is conducted on a current driving road section to obtain an optimal driving path, and speed dynamic planning is automatically conducted on the optimal driving path through a speed optimization algorithm to obtain a rough solution speed curve; and then speed quadratic programming is carried out on the rough solution speed curve in combination with a driver braking decision and a machine braking decision to obtain an optimal speed curve, an optimal driving track is determined according to the optimal driving path and the optimal speed curve, and the vehicle is controlled by taking the optimal driving track as a target. On the basis of automatic dynamic speed planning of the vehicle decision planning system, the braking intention of the driver is recognized in combination with the electroencephalogram signals to carry out speed quadratic planning, corner scenes can be sensed in time, automatic braking can be carried out, and the automatic driving safety of the vehicle is improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle autonomous driving technology, and in particular to a vehicle autonomous driving method, system, electronic device and storage medium. Background Art

[0002] Autonomous driving refers to the ability of motor vehicles to operate autonomously without human intervention through technologies such as artificial intelligence, computer vision, and sensors. Autonomous vehicles rely on the synergy of artificial intelligence, visual computing, radar, monitoring devices, and global positioning systems to safely and automatically operate the vehicle without active human intervention. While the rapid development of perception systems driven by deep learning has made significant progress in the autonomous driving industry, it still faces challenges in its implementation. Visual perception methods are a crucial component of autonomous driving perception systems, detecting the vehicle's surroundings through visual perception. Consequently, a large number of algorithms exist for visual perception tasks related to detecting the vehicle's environment. However, traditional machine learning methods struggle to handle unusual driving conditions, such as those caused by erratic driving by other traffic participants, novel driving conditions not previously seen in datasets, and high-risk driving conditions. These conditions are known as edge cases for autonomous driving.

[0003] These edge scenarios are extreme situations that pose a high safety threat but a low probability of occurrence in real-world applications, and are an essential step in the implementation and verification of autonomous driving. While these situations rarely occur in real-world scenarios, they represent a significant obstacle to the full commercialization of autonomous driving. Currently, research on edge scenarios is limited, and their main characteristics are fragmented scenes, irregular movements of major traffic participants, complex and unpredictable driving environments, high collision costs, and a high risk factor. Applying algorithms designed for normal operating conditions to edge scenarios inevitably leads to problems such as information recognition delays and difficulty in scene processing.

[0004] Current solutions to corner scenarios rely on addressing them one by one. However, corner scenarios are impossible to enumerate. Similar situations have countless variations, making it difficult to address each scenario individually. Furthermore, each scenario carries a high degree of risk. Furthermore, while adding corner scenario samples can improve the ability to handle these scenarios, it can actually harm the neural network, reducing the model's ability to detect normal scenes. Therefore, currently, autonomous vehicles face low safety issues when dealing with corner scenarios. Summary of the Invention

[0005] The main purpose of the embodiments of this application is to propose a vehicle automatic driving method, system, electronic device and storage medium, aiming to improve the safety of vehicle automatic driving.

[0006] To achieve the above objectives, an embodiment of the present application provides a method for autonomous driving of a vehicle, comprising the following steps:

[0007] Collecting the driver's brain electrical signals and identifying the driver's braking decision in real time based on the brain electrical signals;

[0008] Perform path planning for the current driving section to obtain the optimal driving path;

[0009] Using a speed optimization algorithm, dynamic speed planning is performed on the optimal driving path to obtain a rough speed curve;

[0010] performing speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain an optimal speed curve;

[0011] An optimal driving trajectory is determined according to the optimal driving path and the optimal speed curve, and the vehicle is controlled with the optimal driving trajectory as a target.

[0012] In some embodiments, the real-time identification of the driver's braking intention based on the EEG signal includes the following steps:

[0013] Using a common space pattern algorithm to extract features from the EEG signal to obtain EEG features;

[0014] A support vector machine algorithm is used to perform feature classification on the EEG features to obtain the driver's braking intention.

[0015] In some embodiments, the process of performing path planning on the current driving section to obtain the optimal driving path includes the following steps:

[0016] Constructing a Frenet coordinate system according to the road centerline of the current driving section, wherein two coordinate axes of the Frenet coordinate system respectively represent a longitudinal distance and a lateral offset along the road centerline;

[0017] With the goal of minimizing the total path cost, dynamic path planning is performed on the path points in the Frenet coordinate system to obtain the minimum cost path, wherein the total path cost includes the road smoothness cost, the obstacle distance cost, and the center distance cost;

[0018] Performing secondary planning on the minimum cost path to obtain the optimal driving path.

[0019] In some embodiments, performing secondary planning on the minimum cost path to obtain the optimal driving path includes the following steps:

[0020] Determining path parameter constraints based on the drivable space where the minimum cost path is located;

[0021] The minimum cost path is optimized according to the path parameter constraints and the first optimization objective to obtain an optimal driving path.

[0022] In some embodiments, the use of a speed optimization algorithm to dynamically plan the speed of the optimal driving path to obtain a rough speed curve includes the following steps:

[0023] Converting the optimal driving path into a representation of longitudinal distance and time relationship;

[0024] Predicting the obstacle's motion trajectory based on the obstacle's motion state information to obtain the obstacle's predicted trajectory;

[0025] a collision time interval determined based on the predicted obstacle trajectory and the optimal driving path, and determining reference speeds corresponding to collision time points and non-collision time points in the optimal driving path, respectively, based on the collision time interval;

[0026] With the goal of minimizing the total speed cost, dynamic speed planning is performed at each time point of the optimal driving path to obtain a rough speed curve, wherein the total speed cost includes a speed difference cost and an acceleration smoothing cost, and the speed difference cost is calculated by the planned speed and the reference speed.

[0027] In some embodiments, performing speed quadratic programming on the rough-solved speed curve in combination with the driver's braking decision and the machine's braking decision to obtain the optimal speed curve includes the following steps:

[0028] The optimal speed curve at the previous time point is used as the speed reference line at the current time point, and the acceleration constraint is generated according to the speed reference line;

[0029] Optimizing the rough speed curve according to the acceleration constraint and the second optimization objective to obtain an optimal speed curve;

[0030] Wherein, when the driver's braking decision is that there is currently a braking intention, the reference speed in the second optimization objective is zero.

[0031] In some embodiments, the second optimization objective is to minimize the quadratic programming cost, the quadratic programming cost is calculated by a quadratic programming cost function, the quadratic programming cost function includes a reference speed cost term, an acceleration cost term, and an acceleration derivative cost term, and optimizing the rough solution speed curve according to the acceleration constraint and the second optimization objective to obtain the optimal speed curve includes the following steps:

[0032] Dynamically adjusting each weight of the quadratic programming cost function according to the driver's braking decision;

[0033] Calculating the quadratic programming cost of the rough solution speed curve according to the adjusted quadratic programming cost function;

[0034] The rough speed curve is optimized according to the quadratic programming cost and the acceleration constraint until the quadratic programming cost of the rough speed curve is minimized, thereby obtaining an optimal speed curve.

[0035] To achieve the above objectives, another aspect of the present application provides an automatic driving system for a vehicle, comprising:

[0036] The first module is used to collect the driver's EEG signals and identify the driver's braking decision in real time based on the EEG signals;

[0037] The second module is used to plan the path of the current driving section and obtain the optimal driving path;

[0038] The third module is used to use a speed optimization algorithm to perform dynamic speed planning on the optimal driving path to obtain a rough speed curve;

[0039] A fourth module is configured to perform speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain an optimal speed curve;

[0040] The fifth module is used to determine an optimal driving trajectory based on the optimal driving path and the optimal speed curve, and control the vehicle with the optimal driving trajectory as the target.

[0041] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes an electronic device, which includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the method described in the above embodiment is implemented.

[0042] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in the above embodiment.

[0043] The vehicle automatic driving method, system, electronic device and storage medium proposed in this application collect the driver's EEG signals during the automatic driving process, and identify the driver's braking decisions in real time based on the EEG signals. Then, the path planning is performed on the current driving section to obtain the optimal driving path. The speed optimization algorithm is used to automatically perform dynamic speed planning on the optimal driving path to obtain a rough speed curve. The rough speed curve is then combined with the driver's braking decision and the machine's braking decision to perform secondary speed planning to obtain the optimal speed curve. The optimal driving trajectory is determined based on the optimal driving path and the optimal speed curve, and the vehicle is controlled with the optimal driving trajectory as the target. Based on the automatic dynamic speed planning of the vehicle decision-making planning system, this application combines EEG signals to identify the driver's braking intention to perform secondary speed planning, so that it can timely perceive corner scenes and automatically brake, thereby improving the safety of vehicle automatic driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flowchart of the vehicle automatic driving method provided by an embodiment of the present application;

[0045] Figure 2 This is a schematic diagram of an experimental corner scene provided in an embodiment of the present application;

[0046] Figure 3 This is a schematic diagram of the training process of the intent classification algorithm provided in the embodiment of the present application;

[0047] Figure 4 This is a feature extraction flow chart of the common space pattern algorithm provided in an embodiment of the present application;

[0048] Figure 5 Schematic diagram of the optimal hyperplane of the support vector machine provided in the embodiment of the present application;

[0049] Figure 6 is a schematic diagram of the classification results of braking intentions of different subjects provided in an embodiment of the present application;

[0050] Figure 7 This is a flow chart of the path dynamic planning method provided by an embodiment of the present application;

[0051] Figure 8 This is a flowchart of the path quadratic planning solution provided by the embodiment of the present application;

[0052] Figure 9 This is a schematic diagram of the human-machine collaborative speed planning concept provided by an embodiment of the present application;

[0053] Figure 10 It is the speed dynamic programming ST diagram provided in the embodiment of the present application;

[0054] Figure 11 This is a flow chart of the speed dynamic planning method provided in an embodiment of the present application;

[0055] Figure 12 Schematic diagram of the ST rough solution velocity curve and its search area provided in an embodiment of the present application;

[0056] Figure 13 This is a flowchart of solving the quadratic programming of human-machine collaborative speed provided by an embodiment of the present application;

[0057] Figure 14 This is a schematic diagram of the overall trajectory planning concept for autonomous driving of a vehicle provided in an embodiment of the present application;

[0058] Figure 15 This is a schematic diagram of a corner scenario test case provided by an embodiment of the present application;

[0059] Figure 16 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0061] It should be noted that although the system is divided into functional modules and the flowcharts illustrate a logical sequence, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowcharts. The terms "first," "second," and so on in the specification, claims, and drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0063] First, let’s analyze some of the terms used in this application:

[0064] Brain Computer Interface (BCI) is a multidisciplinary technology that collects EEG signals from the brain through an EEG cap, interprets and classifies the EEG signals based on brain science and pattern recognition technologies, and completes direct communication between the brain and external devices. It can extract the driver's braking intention and achieve the purpose of controlling external devices.

[0065] The Frenet coordinate system is a dynamic local coordinate system based on the road centerline and is widely used in the fields of autonomous driving trajectory planning and robotic motion control. The two coordinate axes of the Frenet coordinate system are defined as the curve distance from the starting point to the vehicle projection point along the road centerline reference line, representing the longitudinal position, and the normal distance from the vehicle center of mass to the reference line projection point, representing the lateral offset. The origin of the coordinate system moves dynamically with the vehicle position, with the tangent and normal lines of the current point of the reference line as the coordinate axes. The Frenet coordinate system usually uses the lane centerline as the baseline and generates a smooth reference line through discrete point sampling and optimization algorithms (such as Apollo's scattered point smoothing).

[0066] The SL graph is a representation in the Frenet coordinate system that decomposes the vehicle path into the longitudinal distance (S) and lateral offset (L) along the road centerline.

[0067] SL graphs are commonly used in path planning to simplify the representation of a vehicle's position relative to the road. ST graphs represent the relationship between time (T) and longitudinal distance (S) and are commonly used in speed planning to ensure that the vehicle follows a predetermined spatiotemporal trajectory, avoid collisions, and optimize driving efficiency.

[0068] The embodiments of the present application provide a vehicle autonomous driving method, system, electronic device and storage medium, aiming to improve the safety of vehicle autonomous driving.

[0069] The vehicle automatic driving method, system, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the vehicle automatic driving method in the embodiments of the present application is described.

[0070] The vehicle automatic driving method provided in the embodiment of the present application relates to the field of vehicle automatic driving technology. The vehicle automatic driving method provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server side, and can also be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, or as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the vehicle automatic driving method, etc., but is not limited to the above forms.

[0071] Figure 1 This is an optional flowchart of the vehicle automatic driving method provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.

[0072] Step S101, collecting the driver's EEG signal and identifying the driver's braking decision in real time based on the EEG signal;

[0073] Step S102, performing path planning on the current driving section to obtain the optimal driving path;

[0074] Step S103: Using a speed optimization algorithm, dynamically plan the speed of the optimal driving path to obtain a rough speed curve;

[0075] Step S104, performing speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain the optimal speed curve;

[0076] Step S105 , determining an optimal driving trajectory according to the optimal driving path and the optimal speed curve, and controlling the vehicle with the optimal driving trajectory as a target.

[0077] In steps S101 to S105 shown in the embodiment of the present application, during the automatic driving process, the vehicle collects the driver's EEG signals, and identifies the driver's braking decisions in real time based on the EEG signals. Path planning is then performed on the current driving section to obtain the optimal driving path. A speed optimization algorithm is used to automatically perform dynamic speed planning on the optimal driving path to obtain a rough speed curve. The rough speed curve is then subjected to secondary speed planning based on the driver's braking decisions and the machine's braking decisions to obtain the optimal speed curve. The optimal driving trajectory is determined based on the optimal driving path and the optimal speed curve, and the vehicle is controlled with the optimal driving trajectory as the target. Based on the automatic dynamic speed planning of the vehicle perception system, the present application also combines EEG signals to identify the driver's braking intentions for secondary speed planning, thereby enabling timely perception of corner scenes and automatic braking, thereby improving the safety of the vehicle's automatic driving.

[0078] According to some embodiments of the present application, the real-time identification of the driver's braking intention based on the EEG signal in step S101 may include, but is not limited to, the following steps:

[0079] Step S201, using a common space pattern algorithm to extract features from the EEG signal to obtain EEG features;

[0080] Step S202 : Using a support vector machine algorithm to perform feature classification on the EEG features to obtain the driver's braking intention.

[0081] Specifically, the co-spatial pattern algorithm and the support vector machine algorithm are both machine learning algorithms. Before using machine learning algorithms to identify intent from EEG signals, the machine learning model needs to be trained first. The training process is described as follows:

[0082] First, we designed an offline simulated driving scenario, extracted the driver's emergency braking intention based on EEG signals, and provided a braking intention detection model for each subject for the online experiment. For example, please refer to Figure 2 The experimental corner scene diagram is shown below. The experimental scene description is as follows:

[0083] An autonomous vehicle is driving along a street. On the right side of the vehicle, there is a house blocking the road. The autonomous vehicle is driving normally on the road next to the house. Suddenly, a person crosses the road. Before appearing on the road, the person is completely in the blind spot of the vehicle. Figure 2 As shown in (a) and (c) in . Then, as Figure 2 As shown in (b), a pedestrian suddenly walks out from the gap between the houses next to it. At this time, the subjects will have the intention to brake due to the sudden appearance of the pedestrian, and are required to step on the brake pedal as the event trigger time record.

[0084] In normal driving scenarios, there were no pedestrians suddenly crossing the road, and the subjects were asked to sit still. Before the formal experiment, all subjects underwent a pre-experiment to familiarize themselves with the driving tasks of this experiment (including normal driving + emergency braking conditions). Normal driving and emergency driving experiments were conducted in a random order, so the subjects did not know in advance whether or when an emergency might occur. In offline experiments, there was only one or no emergency in each offline experiment. The reason why an emergency only occurs once is that in actual driving, the probability of an emergency is very low. If the frequency of emergency is set high, the subjects will get used to the occurrence of emergencies, and then the subjects' alertness to crises may also decrease. During the experiment, the subjects need to pay attention to the driving scene on the screen.

[0085] When an emergency occurs, the subject needs to perform emergency braking and step on the brake pedal. The driving simulator will record the time when the subject steps on the brake pedal. Based on this time node, the EEG data 1 second before stepping on the brake pedal is extracted as the emergency braking intention trigger data, and the 4 seconds of time data from the 2nd to the 5th second before stepping on the brake pedal is extracted as the normal driving data. Therefore, 5 seconds of data will be extracted for each braking intention experiment.

[0086] Please refer to Figure 3 The following figure shows the training process of the intent classification algorithm, which performs preprocessing operations after obtaining the raw EEG data:

[0087] 1. Electrode Selection: The human brain is complex, with different brain regions corresponding to distinct functions. Braking intent is associated with human perception, decision-making, and movement, so it's necessary to extract signals from relevant brain regions for braking intent classification. Finally, ten electrodes, F3, Fz, F4, C3, Cz, C4, P3, P4, O1, and O2, which are associated with braking intent, were selected for data collection.

[0088] ② Filtering: There is a lot of noise and interference in reality, which will affect the quality of EEG signals. Currently, filtering methods within 60 Hz are often used for data filtering of EEG braking intention signals. Therefore, in this embodiment, band-pass filtering from 1 - 45 Hz is used to reduce the impact of noise on EEG data.

[0089] ③ Segmenting offline experimental data: During the offline experiment of braking intention detection, the driving simulator records the time when the subject steps on the brake pedal. Therefore, the time when the subject generates the EEG braking intention signal can be inferred, and the EEG data before this time point is collected. Finally, each EEG signal segment is cut into several data segments with a sample length of 1 s for downstream module processing.

[0090] ④ Labeling experimental data: There are two states in the experiment, namely the normal driving state and the emergency braking state. So after data segmentation, each second of EEG data should have a label corresponding to the driving state ("1" and "2", where "1" represents the normal driving state and "2" represents the emergency braking state). Using the recorded brake pedal signal as the end point, the EEG signal 1 s before this is intercepted, and the segmentation and labeling steps are completed. The original EEG data in the channels×times format is converted into the channels×times×trials format after segmentation, where channels represent the number of collected channels, times represent the number of data sampling points, and trials represent the number of normal driving and emergency braking trials.

[0091] The preprocessed raw EEG data uses the Common Spatial Pattern (CSP) algorithm to extract its spatial domain features, and then the extracted features are substituted into the Support Vector Machines (SVM) feature classification algorithm for prediction. The structure of the braking intention detection algorithm is as Figure 3 shown.

[0092] CSP is a spatial domain feature extraction method widely used in binary classification of EEG data. The feature extraction process of the common spatial pattern algorithm is as Figure 4 shown. Assume X1 and X2 are the normal driving state and the emergency braking state, and the dimensions of X1 and X2 are both C×T, where C is the number of channels and T represents the total number of data in a single channel, and C < T. Using the method of mixed sources to describe the EEG signals of the two driving states, then X1 and X2 can be expressed as:

[0093]

[0094] where, O M represents the common source signal of the normal driving state and the emergency braking state, Oi Denote the source signals unique to the two states. O1 is composed of m1 sources, and O2 is composed of m2 sources. Similarly, W1 and W2 are composed of m1 and m2 common spatial patterns associated with X1 and X2, respectively. W1 and W2 denote the weights of the different components, respectively. In summary, the difference between the normal driving and emergency braking signals stems from the difference between W1O1 and W2O2. The principle of the CSP method is to use the difference between W1O1 and W2O2 to classify the two signals. The specific algorithm is as follows.

[0095] Combine the two signals, and then solve the covariance matrices R1 and R2, and let the sum of the two signals be the mixed space covariance matrix R:

[0096] R=R1+R2; (2)

[0097] Decompose the eigenvalues of R into:

[0098] R=UλU T ; (3)

[0099] Where U is the decomposed vector, and λ is the eigenvalue diagonal matrix corresponding to U.

[0100] Then the whitening matrix P is:

[0101]

[0102] The source components are further separated, and the covariance matrices of the normal driving and emergency braking signals are whitened using P to obtain S1 and S2. The eigenvalues are then decomposed to obtain the eigenvectors B1 and B2 of S1 and S2, and the corresponding eigenvalue diagonal matrices λ1 and λ2. B1, B2 and λ1, λ2 satisfy the following: B1 = B2 and λ1 + λ2 = E.

[0103]

[0104] The optimal spatial filter W is constructed based on the matrix P and the vector B (B = B1 = B2), and then W is used to obtain the projections Z1 and Z2 of the normal driving state and emergency braking state signals, as shown in the following formula:

[0105] W=B T P; (7)

[0106] Z1=WX1,Z2=WX2; (8)

[0107] The above covariance matrix of Z1 and Z2 is obtained In order to make the difference between Z1 and Z2 more obvious, the eigenvalues and filters are changed so that one type of features is sorted in descending order and the eigenvalue matrix of the other type is sorted in ascending order.

[0108] Let the test set be X i , use the spatial filter W to project the EEG signal and obtain the covariance matrix Z i (Z i =WX i ), obtain the EEG signal feature f P :

[0109]

[0110] Please continue to refer to Figure 4 , using support vector machine (SVM) as a feature classifier. SVM is a commonly used supervised binary classification algorithm. Its basic principle is to find the best plane to separate the two types of sample data points so that the two types of samples in the sample space can be separated on both sides of the plane. Finally, the problem is resolved into a quadratic programming problem. The plane is called the optimal hyperplane. The optimal hyperplane of the support vector machine is as follows Figure 5 As shown in the figure, the circular and triangular markers are two types of samples, L separates the two types of samples, L1 and L2 are parallel to L, and the distance between L1 and L2 is the classification interval.

[0111] Since the application is binary classification, the given set D = {(x1, y1), (x2, y2) ... (x m ,y m )},y1={+1,1}, where D is the sample data. In the sample plane, the hyperplane can be described by the following method:

[0112] w T x+b=0; (10)

[0113] Where w is the normal vector and b is the displacement term. The distance between any point x in the sample and the hyperplane can be obtained as follows:

[0114]

[0115] Assuming that the hyperplane (w, b) can correctly classify the data points in the sample, for (x i ,y i )∈D has:

[0116]

[0117] The point closest to (w, b) is the support vector. The sum of the closest distances between the two data samples and (w, b) is:

[0118]

[0119] Taking the maximum value of γ can maximize the interval of the hyperplane. Therefore, we need to make ||w|| -1Taking the maximum value is equivalent to making ||w|| 2 Taking the minimum value, change the objective to:

[0120]

[0121] Construct the Lagrangian function as follows:

[0122]

[0123] After solving α, we get w and b, which is the model:

[0124]

[0125] The final decision function is:

[0126] k=sign(w T x i +b); (17)

[0127] The choice of kernel function has a significant impact on the classification performance of the SVM algorithm. Choosing an appropriate kernel function can effectively map data into a high-dimensional space, thereby improving classification results. This example performs a grid search on the polynomial order and penalty coefficient, determines the optimal combination by matching the two parameters, and uses the parameter with the lowest error rate using a five-fold cross-validation.

[0128] Take the offline EEG data of 5 subjects and represent them in 5 groups. Use the CSP-based feature extraction algorithm and the SVM-based feature classification algorithm to divide the EEG data into the data set and the test set in a ratio of 4:1. Figure 6 The figure shows the braking intention classification results for different subjects. It shows that all five subjects had distinct characteristics of emergency braking intention. The feature values extracted from the training set were substituted into the SVM model to train each subject. The training model was then substituted into the test set data for prediction and classification. The braking intention classification accuracy for most subjects was high, with an average accuracy of 94%, indicating strong recognition.

[0129] After building a machine learning-based intent classification algorithm, it can be applied to real-world driving scenarios. In these scenarios, the driver's EEG signals are collected in real time. These signals are then processed and fed into the intent classification algorithm to determine the driver's braking decision, which indicates whether the driver currently intends to brake. Subsequently, human driving experience can be leveraged to enhance the autonomous vehicle's ability to respond to crises, and human-machine collaborative motion planning methods can be used to improve the safety of smart vehicles.

[0130] In step S102 of some embodiments, during the autonomous vehicle planning process, motion planning is decoupled into path planning and velocity planning based on the Frenet coordinate system. The planning process can utilize a combination of dynamic programming and quadratic programming to construct a cost function and find the optimal solution for the current driving environment. The autonomous vehicle driving system first performs path planning for the current driving section to determine the optimal driving path.

[0131] According to some embodiments of the present application, step S102 may include but is not limited to the following steps:

[0132] Step S301: constructing a Frenet coordinate system based on the road centerline of the current driving section, where the two coordinate axes of the Frenet coordinate system represent the longitudinal distance and the lateral offset along the road centerline, respectively;

[0133] Step S302: Dynamically plan the path points in the Frenet coordinate system with the goal of minimizing the total path cost to obtain the minimum cost path, where the total path cost includes the road smoothing cost, the obstacle distance cost, and the center distance cost;

[0134] Step S303: perform secondary planning on the minimum cost path to obtain the optimal driving path.

[0135] For details, please refer to Figure 7 The dynamic path planning flowchart shown here constructs a Frenet coordinate system based on the road centerline of the current driving section. Based on the path sampling points in the Frenet coordinate system, a dynamic programming method is used to open up a drivable area in the SL graph to obtain a rough solution for the path planning, namely the minimum cost path. Assuming that the planned path in the Frenet coordinate system is represented by l = f(s), the cost function used to calculate the total path cost is as follows:

[0136] C total (f(s))=C smooth (f)+C obs (f)+C guidance (f); (18)

[0137] Among them, C total (f(s)) is the total cost of the path, which consists of three parts, C smooth (f) is the road smoothing cost, C obs (f) is the obstacle distance cost, C guidance (f) is the reference line distance cost.

[0138] The road smoothing cost is expressed as:

[0139] C smooth (f)=w1∫(f′(s))2 ds+w2∫(f″(s)) 2 ds+w3∫(f″′(s)) 2 ds; (19)

[0140] Among them, f′(s) represents the heading angle of the vehicle in the natural coordinate system, that is, the angle between the front of the vehicle and the tangent direction of the road; f″(s) represents the curvature of the road; f″′(s) represents the derivative of the road curvature.

[0141] The obstacle cost is expressed as:

[0142]

[0143] Among them, C nudge With (dd c ) increases and decreases. The closer to the safe area, the greater the cost, and the farther away from the safe distance, the smaller the cost; d c Indicates the reserved safety distance; C collision is the collision cost within the safe area, which is set to infinity, that is, all paths within the safe area are infeasible. n Obstacles outside the distance are not considered. nudge and C collision The processing is as follows:

[0144]

[0145] The reference line distance cost is determined by the distance between the planned path and the reference line. The reference line is the center line of the road, and it is set to g(s). The reference line distance cost is expressed as:

[0146] C guidance (f) = f(f(s) - g(s)) 2 ds; (22)

[0147] After obtaining the total cost, the minimum cost path between discrete path points can be calculated based on the dynamic programming method.

[0148] According to some embodiments of the present application, step S303 may include but is not limited to the following steps:

[0149] Step S401, determining path parameter constraints based on the drivable space where the minimum cost path is located;

[0150] Step S402 : Optimizing the minimum cost path according to the path parameter constraints and the first optimization objective to obtain the optimal driving path.

[0151] Specifically, after the minimum cost path is obtained through the path dynamic programming method, the minimum cost path and the drivable space where the path is located are used as the input of the secondary programming. Figure 8 The path quadratic programming solution flowchart shown in the figure optimizes the path based on the rough solution and drivable area provided by dynamic programming and the objective function with linear constraints based on the quadratic programming method, and finally obtains the optimal driving path (i.e., the optimal path) for the current environment.

[0152] In the SL graph, the reference line provides a rough solution to the obstacle bypass (i.e., the minimum cost path). The minimum cost path is further optimized during the path quadratic planning process. The first optimization goal is to minimize the following C s (f) Function:

[0153]

[0154] Among them, g(s) is the result of dynamic programming; f′(s), f″(s), and f″(s) are all path parameters, corresponding to the vehicle heading angle, path curvature, and the derivative of the path curvature, respectively.

[0155] Among them, the quadratic programming constrains f′(s), f″(s), and f″′(s) through constraint boundary conditions and dynamic feasibility to constrain the optimal driving path to the drivable space where the minimum cost path is located. For example, the optimal driving path boundary is constrained in the following way:

[0156] l min ≤l≤l max ; (twenty four)

[0157] Among them, l min is the lower bound of the lane line constraint, l max is the upper bound of the lane constraint.

[0158] When a vehicle faces a static obstacle, the above path planning method can plan an optimal obstacle-avoiding driving path that conforms to vehicle dynamics and is smooth.

[0159] In some embodiments, in steps S103 to S104, under urban conditions, due to unpredictable factors such as pedestrians and other vehicles, as well as the presence of corner scenes such as blind spots, the prediction module of the autonomous driving system cannot accurately judge the behavior of other traffic participants, and the speed decision-making method based on dynamic planning is highly dependent on accurate predictions, and therefore cannot adapt to corner scenes. Based on this, the embodiment of the present application combines the light decision of machine speed planning with the heavy decision of human intention through a method of human-machine collaborative speed planning, uses the result of autonomous driving dynamic planning as the light decision, and uses the braking intention corresponding to the human EEG signal as the heavy decision, and combines the human driver's advanced cognitive ability of crisis with the machine driver's speed planning based on optimization theory to enhance driving safety.

[0160] Please refer to Figure 9 The schematic diagram of the human-machine collaborative speed planning concept shown in the figure shows that in the human-machine collaborative speed control, the machine driver and the human driver serve as joint decision-makers for the longitudinal speed planning of the vehicle. Specifically, for the perception module, the machine driver relies on sensors such as lidar and cameras to perceive the environment, and the human driver relies on vision to perceive the driving environment. The visual signals are then processed by the brain to make decisions. The human driver can cooperate with the machine driver to perceive the environment. For the decision-making planning module, speed planning is performed through a decision-making planning model (i.e., speed optimization algorithm) based on the optimization theory of the machine driver. Then, the perception ability of the human driver is further combined, and the driver's braking decision is identified by using his EEG signal to perform secondary speed planning. This serves as supervision and enhancement of the machine driver's speed planning, allowing the machine driver to complement the advantages of the human driver and obtain the optimal speed curve in emergency situations. This improves the machine driver's ability to respond to corner scenes where perception and decision-making are difficult, thereby providing a guarantee for the safety of autonomous driving.

[0161] According to some embodiments of the present application, step S103 may include but is not limited to the following steps:

[0162] Step S501, converting the optimal driving path into a representation of the longitudinal distance and time relationship;

[0163] Step S502: predicting the obstacle's motion trajectory based on the obstacle's motion state information to obtain a predicted obstacle trajectory;

[0164] Step S503, determining the collision time interval determined based on the predicted obstacle trajectory and the optimal driving path, and determining reference speeds corresponding to the collision time point and the non-collision time point in the optimal driving path based on the collision time interval;

[0165] In step S504, with the goal of minimizing the total speed cost, dynamic speed planning is performed at each time point of the optimal driving path to obtain a rough speed curve. The total speed cost includes the speed difference cost and the acceleration smoothing cost. The speed difference cost is calculated by the planned speed and the reference speed.

[0166] Specifically, speed dynamic planning first projects the optimal driving path of the planned path onto the ST graph to convert the optimal driving path into a representation of the longitudinal distance and time relationship. It then determines whether any obstacles will overlap with the current path on the ST graph. The future trajectory of the obstacle is predicted based on the current motion state information of the obstacle. The motion state of the dynamic obstacle is described as:

[0167]

[0168] Among them, h, and where represents the position, velocity, and acceleration of the obstacle at time t, respectively. Assuming the obstacle's state does not change suddenly and it moves regularly within a certain timeframe, the center position of the obstacle at time t can be predicted. Using the obstacle's size as detected by the vehicle's perception system, the potential collision time with the autonomous vehicle can be calculated.

[0169] The ST diagram required for speed dynamic programming is as follows Figure 10 As shown in the figure, the shaded area represents the area where the dynamic obstacle appears in the ST diagram, l represents the width of the obstacle, and the projection of the obstacle on the T-axis represents the time interval during which a collision with the ego vehicle occurs. The goal of dynamic velocity programming is to find the speed curve with the lowest cost, while satisfying constraints such as vehicle dynamics. This allows the autonomous vehicle to avoid dynamic obstacles while maintaining driving comfort.

[0170] The process of speed dynamic programming is as follows Figure 11 As shown in FIG, the steps include generating an ST diagram, designing a speed cost function, and finally performing a dynamic programming solution. Through the speed dynamic programming, a rough solution of the speed curve, a drivable space for secondary speed planning, and acceleration or deceleration avoidance information can be obtained.

[0171] ST rough solution speed curve is as follows Figure 12 As shown in Figure 1, the optimal speed curve of the previous frame will be used as a reference line in the next frame, and a drivable space for acceleration or deceleration decisions will be generated, providing a search area for quadratic programming.

[0172] The speed dynamic planning first needs to determine the starting point of the speed planning, that is, the longitudinal position, speed and acceleration of the current vehicle in the Frenet coordinate system as the starting point of the planning. Secondly, the end point of the current path is used as the end point of the dynamic planning. The speed cost function between the current position of the vehicle and the end point of the path can be used to calculate the total speed cost of the speed curve and find the rough solution of the speed curve with the minimum total speed cost. Assume that the discrete point set is S = (s0, s1, ... s n ), the speed cost function of the embodiment of the present application is:

[0173]

[0174] Among them, V ref represents the reference speed, which is determined by both the road restrictions and the subject's braking intention. The first term represents the cost of the speed difference: when there is no obstacle, the vehicle drives at the designed normal reference speed; when there is an obstacle, the vehicle drives at the designed smaller reference speed. The second and third terms represent the smoothness of the acceleration and its derivative, respectively. The smaller the cost, the higher the ride comfort.

[0175] Finally, in the ST graph, the trajectory with the minimum cost function searched among discrete speed points using the dynamic programming method is the rough solution speed curve (i.e., the dynamic programming minimum cost speed curve).

[0176] According to some embodiments of the present application, speed quadratic programming is performed on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain the optimal speed curve, which may include but is not limited to the following steps:

[0177] Step S601: The optimal speed curve at the previous time point is used as the speed reference line at the current time point, and an acceleration constraint is generated according to the speed reference line;

[0178] Step S602 : Optimize the rough speed curve according to the acceleration constraint and the second optimization objective to obtain the optimal speed curve.

[0179] When the driver's braking decision indicates that there is currently a braking intention, the reference speed in the second optimization objective is zero.

[0180] According to some embodiments of the present application, the second optimization objective is to minimize the quadratic programming cost. The quadratic programming cost is calculated by a quadratic programming cost function. The quadratic programming cost function includes a reference velocity cost term, an acceleration cost term, and an acceleration derivative cost term. Step S602 may include, but is not limited to, the following steps:

[0181] Step S701, dynamically adjusting the weights of the quadratic programming cost function according to the driver's braking decision;

[0182] Step S702, calculating the quadratic programming cost of the rough solution velocity curve according to the adjusted quadratic programming cost function;

[0183] Step S703 , optimizing the rough speed curve according to the quadratic programming cost and the acceleration constraint until the quadratic programming cost of the rough speed curve is minimized, thereby obtaining the optimal speed curve.

[0184] Specifically, after obtaining the rough speed curve, the speed curve is further optimized based on the speed quadratic programming method. Figure 13 As shown in the figure, a method combining light and heavy decision-making is adopted, with the human driver's crisis awareness ability as the heavy decision and the machine driver's speed decision-making plan as the light decision. That is, when the driver's braking intention is detected, rules are formulated to give priority to the human driver's intention; when the driver's braking intention is not detected, the machine driver's decision-making plan results are given priority.

[0185] The specific quadratic programming cost function is as follows:

[0186]

[0187] Among them, S′ ref is the reference speed: when a braking intention is detected, the reference speed is 0; when no braking intention is detected, the reference speed is the speed curve searched by dynamic programming. The three costs w1, w2, and w3 represent the reference speed cost weight, the acceleration cost weight, and the smoothness of the acceleration derivative (Jerk), respectively. The smaller the Jerk, the better the passenger comfort. Therefore, under normal driving conditions, comfort is more important, and w3 should be set larger. However, in emergency conditions (i.e., when the driver's braking decision indicates an intention to brake), safety is more important, so w3 should be set smaller and w1 should be larger. Therefore, a human-machine collaborative approach is used to adjust the weight coefficients of the quadratic programming cost function. Different weight coefficients w1 and w3 are used for emergency and normal driving conditions. This prioritizes the driver's emergency braking intention, increasing the weight w1 of the recommended speed and decreasing the weight w3 of the Jerk. In the short term, passenger safety is prioritized over ride comfort.

[0188] The constraint boundary of the quadratic programming cost function is expressed as:

[0189]

[0190] The first item is the reversing constraint, which prohibits reverse driving. The second item is the maximum speed limit, which is a fusion of road regulations and the driver's braking intention detection results. The current maximum speed is the minimum of the road signs and the braking intention detection results. The third item is the vehicle dynamics constraint on maximum acceleration and maximum deceleration, which means that the acceleration must meet the vehicle dynamics characteristics. The fourth and fifth items are the position continuity constraints and speed continuity constraints, respectively. The speed curve that takes the extreme value under the conditions that satisfy the constraints is the optimal speed curve.

[0191] In step S105 of some embodiments, the optimal driving path and the optimal speed curve are combined to form the optimal driving trajectory, and the vehicle automatic driving system performs safe driving according to the optimal driving trajectory.

[0192] According to some embodiments of this application, please refer to Figure 14The schematic diagram of the overall trajectory planning concept of the vehicle's autonomous driving is shown. The embodiment of the present application is based on the Frenet coordinate system and decouples motion planning into two parts: path planning and speed planning. A cost function is constructed by combining dynamic programming and quadratic programming to find the optimal solution under the current driving environment. On this basis, human supervision is introduced in the speed planning stage. Human supervision is regarded as an important decision, and the planning method based on optimization theory is regarded as a light decision. The human-machine collaborative method is used for speed planning. Path planning improves the safety of path driving, and human-machine collaborative speed planning improves the ability of autonomous driving vehicles to avoid obstacles at a safe speed when there are dynamic obstacles, and in potential traffic hazard scenarios, the driver's braking intention detection is used to slow down in advance to deal with possible dangerous scenarios. Common dynamic programming solution methods are restricted by sampling density, and it is difficult to obtain the optimal solution for dynamic programming; while the quadratic programming method is restricted by the convex space of the solution range, and it is easy to obtain a local optimal solution. Combining the advantages of dynamic programming and quadratic programming, a method combining dynamic programming and quadratic programming is adopted. Dynamic programming is used to generate the rough solution of the drivable area and path required by the quadratic programming. When seeking the optimal solution, the quadratic programming uses the rough solution of dynamic programming as a reference, reducing the possibility of solving a local optimal solution. Finally, the optimal solution within the drivable area obtained by dynamic programming is solved based on the quadratic programming, the optimal trajectory is obtained, and the vehicle's automatic driving is realized.

[0193] According to some embodiments of the present application, in order to test the performance of the method of the embodiment of the present application, the following three corner scenarios are set up.

[0194] The first scenario is that a pedestrian suddenly appears in the blind spot. Figure 15 In the scenarios (a), (b), and (c), a pedestrian is behind a building and initially in the blind spot of the autonomous vehicle, pushing a cart across the road at a constant speed of 2.8 m / s.

[0195] The second scenario is an unprotected intersection with a blind spot, such as Figure 15 In scenarios (d), (e), and (f), the ego vehicle approaches an unprotected intersection. Another vehicle on the right side of the intersection is stationary when it is more than 30 meters away from the ego vehicle. It then travels at a constant speed of 8 meters per second and arrives at the intersection at the same time as the autonomous vehicle, but does not yield the right of way. Because the gas station blocks the ego vehicle's view, the perception system cannot detect the traffic conditions on the right side of the intersection in time.

[0196] The third situation is that the vehicle on the auxiliary road ahead suddenly changes lanes, such as Figure 15 In scenarios (g), (h), and (i), the ego vehicle is traveling at the default speed. There is another vehicle parked in front of it on the auxiliary road. When the two vehicles approach and are 20 meters apart, the vehicle in front suddenly starts and forcibly merges into the ego vehicle's lane at a speed of 8 m / s.

[0197] In order to reflect the overall control accuracy, safety, and effectiveness under emergency conditions of the method of the embodiment of the present application, the following performance evaluation indicators are selected for evaluation:

[0198] ① Number of collisions: the number of collisions between the vehicle and other vehicles or pedestrians.

[0199] ② Tracking error: the error between the actual driving path and the reference path.

[0200] ③ Driving time: While ensuring safety, driving efficiency is also very important. Record the total time it takes for the machine driver and the human-machine collaborative system to complete the driving task.

[0201] ④ Completion time ratio: Taking the time it takes for the autonomous driving to complete the driving task as the unit time, calculate the ratio of the total time it takes for the subject to complete the driving task to the unit time.

[0202] A simulation test method was used to evaluate the risk perception performance of an intelligent vehicle in a human-machine collaborative mode. Simulation tests were conducted using PreScan, Carsim, and Simulink, analyzing the effectiveness of the method described in this application's embodiments by combining metrics such as the number of collisions, trajectory deviation, and response time. To address the uncertainty of corner scenarios, the intelligent vehicle integrates a perception system, trajectory planning, and optimized control algorithms, while also incorporating recognition of the driver's emergency braking intent to enhance its adaptability to complex traffic environments.

[0203] The method in the embodiments of this application can effectively extract the driver's braking intention in simulation scenarios and integrate it with the autonomous driving system, providing stronger collision avoidance capabilities in corner scenarios under emergency conditions compared to current autonomous driving systems. This technology can improve the safety of unmanned driving and enhance the cognitive capabilities of intelligent vehicles in understanding and predicting traffic environments. Furthermore, this technology has broad application prospects and can be expanded to other intelligent driving assistance systems, further enhancing the intelligence level of electric vehicles.

[0204] The present application also provides an automatic driving system for a vehicle, including:

[0205] The first module is used to collect the driver's EEG signals and identify the driver's braking decision in real time based on the EEG signals;

[0206] The second module is used to plan the path of the current driving section and obtain the optimal driving path;

[0207] The third module is used to use the speed optimization algorithm to dynamically plan the speed of the optimal driving path and obtain a rough speed curve;

[0208] The fourth module is used to perform speed quadratic programming on the rough speed curve by combining the driver's braking decision and the machine's braking decision to obtain the optimal speed curve;

[0209] The fifth module is used to determine the optimal driving trajectory based on the optimal driving path and the optimal speed curve, and control the vehicle with the optimal driving trajectory as the target.

[0210] It can be understood that the contents of the above-mentioned vehicle automatic driving method embodiments are applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those in the above-mentioned vehicle automatic driving method embodiments, and the beneficial effects achieved are also the same as those achieved by the above-mentioned vehicle automatic driving method embodiments.

[0211] The present application also provides an electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for enabling communication between the processor and the memory. When the program is executed by the processor, the aforementioned method for autonomous driving of a vehicle is implemented. The electronic device may be any intelligent terminal, such as a tablet computer or an in-vehicle computer.

[0212] See also Figure 16 , Figure 16 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0213] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0214] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the vehicle automatic driving method of the embodiments of this application;

[0215] Input / output interface 903, used to implement information input and output;

[0216] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0217] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0218] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0219] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned vehicle automatic driving method.

[0220] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0221] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0222] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0223] The system embodiment described above is merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0224] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0225] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0226] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0227] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the above units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0228] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0229] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0230] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store programs.

[0231] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A method for automatic vehicle driving, characterized in that: The following steps are involved: Collecting the driver's brain electrical signals and identifying the driver's braking decision in real time based on the brain electrical signals; Perform path planning for the current driving section to obtain the optimal driving path; Using a speed optimization algorithm, dynamic speed planning is performed on the optimal driving path to obtain a rough speed curve; performing speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain an optimal speed curve; An optimal driving trajectory is determined according to the optimal driving path and the optimal speed curve, and the vehicle is controlled with the optimal driving trajectory as a target.

2. The vehicle automatic driving method according to claim 1, characterized in that: The method of identifying the driver's braking intention in real time based on the EEG signal comprises the following steps: Using a common space pattern algorithm to extract features from the EEG signal to obtain EEG features; A support vector machine algorithm is used to perform feature classification on the EEG features to obtain the driver's braking intention.

3. The vehicle automatic driving method according to claim 1, characterized in that: The process of performing path planning on the current driving section to obtain the optimal driving path includes the following steps: Constructing a Frenet coordinate system according to the road centerline of the current driving section, wherein two coordinate axes of the Frenet coordinate system respectively represent a longitudinal distance and a lateral offset along the road centerline; With the goal of minimizing the total path cost, dynamic path planning is performed on the path points in the Frenet coordinate system to obtain the minimum cost path, wherein the total path cost includes the road smoothness cost, the obstacle distance cost, and the center distance cost; Performing secondary planning on the minimum cost path to obtain the optimal driving path.

4. The vehicle automatic driving method according to claim 3, characterized in that: The performing of secondary planning on the minimum cost path to obtain the optimal driving path includes the following steps: Determining path parameter constraints based on the drivable space where the minimum cost path is located; The minimum cost path is optimized according to the path parameter constraints and the first optimization objective to obtain an optimal driving path.

5. The vehicle automatic driving method according to claim 1, characterized in that: The method of using a speed optimization algorithm to dynamically plan the speed of the optimal driving path to obtain a rough speed curve includes the following steps: Converting the optimal driving path into a representation of longitudinal distance and time relationship; Predicting the obstacle's motion trajectory based on the obstacle's motion state information to obtain the obstacle's predicted trajectory; a collision time interval determined based on the predicted obstacle trajectory and the optimal driving path, and determining reference speeds corresponding to collision time points and non-collision time points in the optimal driving path, respectively, based on the collision time interval; With the goal of minimizing the total speed cost, dynamic speed planning is performed at each time point of the optimal driving path to obtain a rough speed curve, wherein the total speed cost includes a speed difference cost and an acceleration smoothing cost, and the speed difference cost is calculated by the planned speed and the reference speed.

6. The automatic driving method for a vehicle according to claim 5, characterized in that: The method of performing speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain the optimal speed curve includes the following steps: The optimal speed curve at the previous time point is used as the speed reference line at the current time point, and the acceleration constraint is generated according to the speed reference line; Optimizing the rough speed curve according to the acceleration constraint and the second optimization objective to obtain an optimal speed curve; Wherein, when the driver's braking decision is that there is currently a braking intention, the reference speed in the second optimization objective is zero.

7. The vehicle automatic driving method according to claim 6, characterized in that: The second optimization objective is to minimize the quadratic programming cost, the quadratic programming cost is calculated by a quadratic programming cost function, the quadratic programming cost function includes a reference speed cost term, an acceleration cost term, and an acceleration derivative cost term, and the rough solution speed curve is optimized according to the acceleration constraint and the second optimization objective to obtain the optimal speed curve, including the following steps: Dynamically adjusting each weight of the quadratic programming cost function according to the driver's braking decision; Calculating the quadratic programming cost of the rough solution speed curve according to the adjusted quadratic programming cost function; The rough speed curve is optimized according to the quadratic programming cost and the acceleration constraint until the quadratic programming cost of the rough speed curve is minimized, thereby obtaining an optimal speed curve.

8. A vehicle automatic driving system, characterized in that: include: The first module is used to collect the driver's EEG signals and identify the driver's braking decision in real time based on the EEG signals; The second module is used to plan the path of the current driving section and obtain the optimal driving path; The third module is used to use a speed optimization algorithm to perform dynamic speed planning on the optimal driving path to obtain a rough speed curve; A fourth module is configured to perform speed quadratic programming on the rough speed curve in combination with the driver's braking decision and the machine's braking decision to obtain an optimal speed curve; The fifth module is used to determine an optimal driving trajectory based on the optimal driving path and the optimal speed curve, and control the vehicle with the optimal driving trajectory as the target.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for implementing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the method described in any one of claims 1 to 7 are implemented.

10. A storage medium, which is a computer-readable storage medium and is used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of claims 1 to 7.

Images