A car automatic driving system

Abstract

The present invention discloses an automobile automatic driving system, which relates to the field of automatic driving technology. The system includes the following components: a behavior prediction algorithm module: a model is constructed using machine learning and deep learning technologies, and a large amount of historical traffic data is trained and learned to identify the behavior patterns of road users and predict their future behavior intentions and possible actions. The present invention uses machine learning and deep learning technologies through a behavior prediction algorithm module to train and learn a large amount of historical traffic data to identify the behavior patterns of road users, and predict their future behavior intentions and possible actions by analyzing the movement trajectories, speeds and acceleration information of pedestrians, bicycles and other vehicles, and combining traffic signals and road conditions, so that the automatic driving car can respond in advance to avoid potential dangers, thereby greatly improving the safety and reliability of the automatic driving system.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to an automobile autonomous driving system. Background Art

[0002] With the rapid development of computer technology, breakthroughs in artificial intelligence and machine learning have provided strong technical support for the research and development of autonomous driving technology. The automobile autonomous driving system is an intelligent automobile system that achieves unmanned driving through an on-board computer system. This system integrates sensors, high-precision maps, positioning, network communications and lidar information technology, enabling the vehicle to autonomously perceive the environment, make decisions, plan and control execution, and achieve safe and efficient driving.

[0003] However, traditional autonomous driving systems mainly rely on real-time sensor data and environmental perception to make decisions, obtain surrounding environment information through lidar and cameras, and combine advanced algorithms for target detection, tracking and identification. This real-time perception and decision-making method has limitations when dealing with complex traffic scenarios. Due to the large amount of uncertainty and dynamic changes in the traffic environment, real-time perception data can often only reflect the current road conditions, and cannot accurately predict possible changes in the future.

[0004] Therefore, how to develop an automatic driving system for cars that can predict the behavior of other road users and perform rapid information processing and real-time prediction has become an important direction for the current development of automatic driving technology. However, most of the automatic driving systems on the market can only meet some of the requirements of the above characteristics and cannot achieve predictions in complex traffic environments. Therefore, developing an automatic driving system for cars that can comprehensively realize the above characteristics is of great significance to improving the safety and reliability of the automatic driving system. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and provide an automobile automatic driving system. It can use machine learning and deep learning technologies through a behavior prediction algorithm module to train and learn a large amount of historical traffic data to identify the behavior patterns of road users. By analyzing the movement trajectories, speeds and accelerations of pedestrians, bicycles and other vehicles, and combining traffic signals and road conditions, it predicts their future behavioral intentions and possible actions. This prediction capability enables the automatic driving car to respond in advance and avoid potential dangers, thereby greatly improving the safety and reliability of the automatic driving system.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a remote interactive education system, characterized in that the system includes the following components:

[0007] Sensor and environmental perception module: collects information about the surrounding environment in real time through sensors, processes and analyzes the collected data, extracts the location, speed, and direction information of road users, and builds an accurate environmental model based on the processed data to provide data support for predictive driving algorithms;

[0008] Behavior prediction algorithm module: Use machine learning and deep learning technologies to build models, train and learn from a large amount of historical traffic data to identify the behavior patterns of road users, predict their future behavioral intentions and possible actions, and update and optimize the prediction model in real time to adapt to different traffic scenarios and changes in road users;

[0009] Decision-making planning and control module: Based on the output of the behavior prediction algorithm module, it evaluates the potential conflicts and dangers between the autonomous vehicle and other road users, generates the decision-making plan of the autonomous vehicle based on the risk assessment results, converts the decision-making plan into specific control instructions, and adjusts the driving state of the autonomous vehicle through the control execution module;

[0010] Learning and Optimization Module: Responsible for collecting driving data and feedback information of autonomous vehicles, continuously learning and optimizing the predictive driving algorithm. By analyzing data from actual driving, the system can continuously improve the prediction model and enhance the accuracy and reliability of the prediction.

[0011] Furthermore, the specific steps of the sensor and the environment perception module are:

[0012] (1) Data processing: Sensors are used to collect information about the surrounding environment in real time, and the collected data is processed and analyzed to extract the location, speed, and direction information of road users;

[0013] (2) Environmental model construction: Based on the processed data, an accurate environmental model is constructed to provide data support for the predictive driving algorithm.

[0014] Furthermore, the behavior prediction algorithm module includes the following specific steps:

[0015] (1) Data collection and preprocessing: Collect a large amount of historical traffic data, including the movement trajectories of pedestrians, bicycles, and other vehicles, as well as traffic signals and road conditions. Clean, denoise, and annotate the collected data to extract useful features.

[0016] (2) Model training and optimization: Use machine learning and deep learning techniques to train preprocessed data, build prediction models, evaluate the prediction performance of the models, and optimize the prediction effects of the models by adjusting model parameters or trying different algorithms.

[0017] (3) Real-time update: In actual operation, new traffic data is continuously collected and the model is updated in real time to adapt to changes in the traffic environment.

[0018] Furthermore, the decision-making planning and control module includes the following specific steps:

[0019] (1) Risk assessment: Receive the predicted behavior of road users from the behavior prediction algorithm module and assess the potential conflicts and dangers between the autonomous vehicle and other road users based on the prediction results;

[0020] (2) Issuing decision instructions: Based on the risk assessment results, a decision plan for the autonomous vehicle is generated, the decision plan is converted into specific control instructions, and the driving state of the autonomous vehicle is adjusted through the control execution module.

[0021] Furthermore, the learning and optimization module includes the following specific steps:

[0022] (1) Data collection: Collect driving data of the autonomous vehicle, as well as feedback information from sensors and other modules to understand the performance of the autonomous vehicle in actual operation;

[0023] (2) Model adjustment and optimization: Based on the collected data and feedback information, the predictive driving algorithm is adjusted and optimized to improve prediction accuracy and performance.

[0024] Furthermore, the behavior prediction algorithm module uses CNN and LSTM to build an algorithm model, thereby realizing accurate prediction of road user behavior. The formula is: convolution operation is performed on the preprocessed historical information to extract local features. The extraction formula is: in, is the output of the lth convolutional layer at position (i, j), is the weight of the convolution kernel, is the input of the previous layer at the corresponding position, b (l) is a bias term. ReLU activation function is usually used to perform nonlinear transformation on the output of the convolutional layer. The formula is: By performing a pooling operation on the activated feature map, the spatial size of the data is reduced. The formula is: Repeat the combination of convolutional layers, activation function layers, and pooling layers to extract deeper features. The features extracted by CNN are used as the input sequence of LSTM, which is processed through the internal gate and state mechanism to output the hidden state of each time step. The output of the LSTM layer can be used as the input of the subsequent layer for behavior prediction to capture the behavioral changes of road users. In the model training process, the mean square error (MSE) is used as the loss function, that is: Among them, yi Indicates actual behavior. Represents the prediction behavior, N represents the number of samples, and the parameters of the model are continuously iteratively updated through the back propagation algorithm and gradient descent optimizer to minimize the loss function and improve the accuracy of the prediction.

[0025] Furthermore, the decision planning and control module solves the optimal decision planning trajectory through a dynamic programming algorithm, and receives the output of the behavior prediction algorithm module, that is, the prediction result of the future behavior of other road users, and assumes that the prediction result set is P = {p1, p2, ..., p n}, where p i represents the predicted behavior trajectory of the i-th road user. The module evaluates potential conflicts and dangers by calculating the distance and speed difference between the autonomous vehicle and each predicted behavior trajectory. The calculation formula is: Among them, CarPos t represents the position p of the autonomous vehicle at time t i (t) represents the predicted behavior trajectory p i At time t, T represents the time range of the evaluation. After evaluating potential conflicts and dangers, the decision planning and control module uses a dynamic programming-based algorithm to make decisions. Let the state space of the autonomous vehicle be X, the control input space be U, and the cost function be J(x i ,u ∈ ), where x i ∈X represents the current state, u ∈ ∈U represents the control input, and the calculation formula is: J(x i ,u ∈ )=α·Safety(x i ,u ∈ )+β· Among them, α, β, γ are weight coefficients, which can be adjusted according to actual needs. For each time step t, the module calculates the current state x t Starting from the cost function value under all possible control input sequences, the control input sequence that minimizes the cost function is selected as the optimal decision planning trajectory. By iteratively solving the optimal control input at each time step, the optimal decision planning trajectory is finally obtained.

[0026] Furthermore, the learning and optimization module uses a gradient descent algorithm to update the model parameters, calculates the gradient of the loss function with respect to the model parameters, and iteratively updates the parameters according to the learning rate. The calculation formula of the loss function is: Specifically, for each new sample (x i ,y i), calculate the gradient of the loss function with respect to the model parameters, Update the parameters according to the learning rate η: By continuously iterating and updating parameters, the prediction error can be gradually reduced and the prediction accuracy can be improved. The updated model will be deployed in autonomous vehicles for real-time prediction and decision-making. By working in collaboration with other modules, autonomous vehicles can better adapt to changes in the actual traffic environment and improve safety and reliability.

[0027] Furthermore, the sensor and environment perception module uses the target detection algorithm of the YOLO framework to locate and classify the target. Assuming the input image is I and the feature vector extracted by the convolutional neural network is x, the target detection algorithm calculates the score and bounding box coordinates of the target by the following formula: Among them, S represents the score of the target, σ represents the activation function, which is used to map the score to the interval [0,1], and W s represents the weight of the scoring layer, x represents the feature vector, and B represents the coordinates and size of the bounding box, including the coordinates of the center point (x c ,y c ) width w and height h, W b Represents the weight of the bounding box layer. During the training process, W is continuously updated through the back propagation algorithm and gradient descent optimization method. s and W b The weight parameters enable the model to accurately identify the target in the image.

[0028] Furthermore, the system also includes a human-computer interaction interface for displaying prediction results, decision-making planning information, and providing an interface for user control and intervention, so that users can understand the operating status and decision-making process of the autonomous driving vehicle and intervene manually when necessary.

[0029] Compared with the existing technology, the car automatic driving system has the following beneficial effects:

[0030] 1. The present invention uses machine learning and deep learning technologies through a behavior prediction algorithm module to train and learn a large amount of historical traffic data to identify the behavior patterns of road users. By analyzing the movement trajectories, speed and acceleration information of pedestrians, bicycles and other vehicles, and combining traffic signals and road conditions, their future behavior intentions and possible actions are predicted, so that the autonomous driving car can respond in advance and avoid potential dangers, thereby greatly improving the safety and reliability of the autonomous driving system.

[0031] 2. The present invention uses a behavior prediction algorithm module to update and optimize the prediction model in real time to adapt to changes in different traffic scenarios and road users. The module's adaptive and optimization capabilities enable it to maintain a high level of prediction accuracy even in complex and changeable traffic environments, further enhancing the robustness and adaptability of the autonomous driving system.

[0032] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 This is a flowchart of the car's automatic driving system process. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Example 1: Implementation of behavior prediction algorithm module:

[0037] In the behavior prediction algorithm module of the present invention, a prediction model based on a combination of a convolutional neural network (CNN) and a long short-term memory network (LSTM) is used. Figure 1 As shown in the figure, first, a large amount of historical traffic data is collected through the sensor and environment perception module, including the movement trajectory, speed and acceleration information of road users, as well as traffic signals and road conditions. Then, the data is preprocessed, including data cleaning, denoising and labeling, to extract useful features. Next, CNN is used to extract the spatial features of road users, and then LSTM is used to process the time series data to capture the behavioral changes of road users. Convolution operation is performed on the input data to extract local features:

[0038]

[0039] in, is the output of the lth convolutional layer at position (i, j), is the weight of the convolution kernel, is the input of the previous layer at the corresponding position, b (l) is a bias term. ReLU activation function is usually used to perform nonlinear transformation on the output of the convolutional layer. The formula is:

[0040]

[0041] By performing a pooling operation on the activated feature map, the spatial size of the data is reduced. The formula is:

[0042]

[0043] Repeat the combination of convolutional layers, activation function layers, and pooling layers to extract deeper features. The features extracted by CNN are used as the input sequence of LSTM, and the input gate is used to determine which feature information needs to be remembered by the LSTM unit:

[0044] i t =σ(W xi x t +W hi h t-1 +b i ),

[0045] Among them, x t is the input at the current moment, h t-1 is the hidden state at the previous moment, W xi and W hi is the weight matrix, b i is the bias term, σ is the activation function, and the forget gate determines which information needs to be forgotten from the LSTM unit:

[0046] c t =f t ⊙c t-1 +i t ⊙tanh(W xc x t +W hc h t-1 +b c ),

[0047] Among them, ⊙ represents element-by-element multiplication, and the output gate determines the output of the LSTM unit based on the new cell state:

[0048] o t =σ(W xo x t +W ho h t-1 +b o ),

[0049] h t =o t ⊙tanh(c t )

[0050] The output of the LSTM is fused with the features extracted by the CNN and then passed to one or more fully connected layers to produce the final prediction output:

[0051] Predicted Output=FC(Concat(h t ,h)),

[0052] Among them, Concat represents the connection operation. Through the above steps, a predictive driving algorithm model combining CNN and LSTM can be constructed. In the model training process, the mean square error (MSE) is used as the loss function, that is:

[0053]

[0054] Among them, y i Indicates actual behavior. Represents the prediction behavior, N represents the number of samples, and the parameters of the model are continuously iteratively updated through the back propagation algorithm and gradient descent optimizer to minimize the loss function and improve the accuracy of the prediction.

[0055] In summary, through the above steps, the present invention uses CNN and LSTM to construct a behavior prediction algorithm model for road user behavior prediction in an autonomous driving system. The model can extract deep features of input data and consider the dependencies of time series, thereby achieving accurate prediction of the behaviors of other road users. Through continuous learning and optimization, the model can adapt to changes in different traffic scenarios and road users, thereby improving the safety and reliability of the autonomous driving system.

[0056] Example 2: Implementation of decision-making planning and control module:

[0057] In the autonomous driving system, the decision planning and control module is responsible for generating and executing corresponding decision planning and control instructions according to the prediction results of the behavior prediction algorithm module. The decision planning and control module first receives the output of the behavior prediction algorithm module, that is, the prediction results of the future behavior of other road users. Suppose the prediction result set is P = {p1, p2, ..., p n}, where p i represents the predicted behavior trajectory of the i-th road user. The module evaluates potential conflicts and dangers by calculating the distance and speed difference parameters between the autonomous driving car and each predicted behavior trajectory. Specifically, the following formula can be used to calculate the distance between the autonomous driving car and the predicted behavior trajectory p iThe minimum distance d between i , the calculation formula is:

[0058]

[0059] Among them, CarPos t represents the position of the autonomous vehicle at time t, p i (t) represents the predicted behavior trajectory p i At time t, T represents the time range of the evaluation. After evaluating potential conflicts and dangers, the decision planning and control module uses a dynamic programming-based algorithm to make decisions. Let the state space of the autonomous vehicle be X, the control input space be U, and the cost function be J(x i ,u ∈ ), where x i ∈X represents the current state, u ∈ ∈U represents the control input, and the calculation formula is:

[0060]

[0061] Among them, α, β, γ are weight coefficients, which can be adjusted according to actual needs. The decision planning and control module solves the optimal decision planning trajectory through the dynamic programming algorithm. Specifically, for each time step t, the module calculates the optimal decision planning trajectory from the current state x t Starting from, the cost function value under all possible control input sequences, and selecting the control input sequence that minimizes the cost function as the optimal decision planning trajectory. Mathematically, this can be expressed as solving the following optimization problem:

[0062]

[0063] Among them, T represents the time range of planning. By iteratively solving the optimal control input of each time step, the optimal decision planning trajectory is finally obtained. After generating the optimal decision planning trajectory, the decision planning and control module converts the trajectory into specific control instructions, and adjusts the driving state of the autonomous driving vehicle through the control execution module. The control instructions include acceleration and steering angle, which are used to guide the autonomous driving vehicle to drive according to the decision planning trajectory. The module calculates the control instructions for each time step according to the decision planning trajectory and sends them to the control execution module for execution. The control execution module is responsible for executing the control instructions and adjusting the driving state of the autonomous driving vehicle by adjusting the throttle, brake and steering systems of the autonomous driving vehicle. At the same time, the module will continuously monitor the real-time status of the autonomous driving vehicle and adjust the control instructions as needed to ensure that the autonomous driving vehicle can respond to road changes and behavioral changes of other road users in real time.

[0064] In summary, through the decision-making planning and control module of this embodiment, the autonomous driving system can use mathematical formulas and algorithms to generate and execute corresponding decision-making planning and control instructions based on the prediction results of the behavior prediction algorithm module, thereby achieving safe interaction with other road users and improving the safety and reliability of autonomous driving.

[0065] Example 3: Implementation of the learning and optimization module:

[0066] The learning and optimization module is an important part of the autonomous driving system. It is responsible for collecting the driving data and feedback information of the autonomous driving car, and continuously learning and optimizing the predictive driving algorithm to improve the performance and accuracy of the algorithm. First, the driving data and feedback information of the autonomous driving car are collected. The driving data includes the actual driving trajectory, speed and acceleration information, while the feedback information comes from the real-time data of sensors and other modules. Suppose the driving data set is D = {(x1, y1), (x2, y2),…, (x n ,y n )}, where x i represents the characteristics of the i-th sample (such as road conditions, traffic signals), y i The module preprocesses the collected data, including data cleaning, feature extraction, and standardization operations, so that the subsequent algorithms can better utilize the data. Next, the learning and optimization module uses the collected data to evaluate the predictive driving algorithm and calculate the difference between the predicted results and the actual behavior. Suppose the model of the predictive driving algorithm is f(x i ; θ), where x represents the input feature and θ represents the model parameter. For each sample (x i ,y i ), the model will output a predicted value By calculating the gradient of the loss function with respect to the model parameters and iteratively updating the parameters according to the learning rate, the loss function is calculated as follows:

[0067]

[0068] At the same time, the learning and optimization module uses an online learning algorithm to update the model parameters. The online learning algorithm can process new data in real time and update the model parameters to adapt to the changing environment. In this embodiment, the gradient descent algorithm is used as the online learning algorithm. The gradient descent algorithm calculates the gradient of the loss function with respect to the model parameters and iteratively updates the parameters according to the learning rate. Specifically, for each new sample (x i ,y i ), calculate the gradient of the loss function with respect to the model parameters,

[0069]

[0070] Update the parameters according to the learning rate η:

[0071]

[0072] By continuously iterating and updating parameters, the prediction error can be gradually reduced and the prediction accuracy can be improved. The updated model will be deployed in autonomous vehicles for real-time prediction and decision-making. By working in collaboration with other modules, autonomous vehicles can better adapt to changes in the actual traffic environment and improve safety and reliability.

[0073] In summary, through the learning and optimization module of this embodiment, we can use online learning algorithms and other optimization technologies to continuously learn and optimize the predictive driving algorithm and improve the performance and accuracy of the autonomous driving system, which helps autonomous driving vehicles better adapt to complex traffic environments and improve driving safety and comfort.

[0074] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An automatic driving system for a car, characterized in that: The system consists of the following components: Sensor and environmental perception module: collects information about the surrounding environment in real time through sensors, processes and analyzes the collected data, extracts the location, speed, and direction information of road users, and builds an accurate environmental model based on the processed data to provide data support for predictive driving algorithms; Behavior prediction algorithm module: Use machine learning and deep learning technologies to build models, train and learn from a large amount of historical traffic data to identify the behavior patterns of road users, predict their future behavioral intentions and possible actions, and update and optimize the prediction model in real time to adapt to different traffic scenarios and changes in road users; Decision-making planning and control module: Based on the output of the behavior prediction algorithm module, it evaluates the potential conflicts and dangers between the autonomous vehicle and other road users, generates the decision-making plan of the autonomous vehicle based on the risk assessment results, converts the decision-making plan into specific control instructions, and adjusts the driving state of the autonomous vehicle through the control execution module; Learning and optimization module: responsible for collecting driving data and feedback information of self-driving cars, continuously learning and optimizing predictive driving algorithms. By analyzing data from actual driving, the system can continuously improve the prediction model and enhance the accuracy and reliability of predictions. in, The specific steps of the behavior prediction algorithm module are: (1) Data collection and preprocessing: Collect a large amount of historical traffic data, including the movement trajectories of pedestrians, bicycles, and other vehicles, as well as traffic signals and road conditions. Clean, denoise, and annotate the collected data to extract useful features. (2) Model training and optimization: Use machine learning and deep learning techniques to train preprocessed data, build prediction models, evaluate the prediction performance of the models, and optimize the prediction effects of the models by adjusting model parameters or trying different algorithms. (3) Real-time update: In actual operation, new traffic data is continuously collected and the model is updated in real time to adapt to changes in the traffic environment; as well as, The behavior prediction algorithm module uses CNN and LSTM to build an algorithm model to achieve accurate prediction of road user behavior. The formula is: convolution operation is performed on the preprocessed historical information to extract local features. The extraction formula is: in, is the output of the lth convolutional layer at position (i, j), is the weight of the convolution kernel, is the input of the previous layer at the corresponding position, b (l) is the bias term, and the ReLU activation function is used to perform a nonlinear transformation on the output of the convolutional layer. The formula is: By performing a pooling operation on the activated feature map, the spatial size of the data is reduced. The formula is: Repeat the combination of convolutional layers, activation function layers, and pooling layers to extract deeper features. The features extracted by CNN are used as the input sequence of LSTM, which is processed through the internal gate and state mechanism to output the hidden state of each time step. The output of the LSTM layer can be used as the input of the subsequent layer for behavior prediction to capture the behavioral changes of road users. In the model training process, the mean square error (MSE) is used as the loss function, that is: Among them, y i Indicates actual behavior. Represents the prediction behavior, N represents the number of samples, and the parameters of the model are continuously iteratively updated through the back propagation algorithm and gradient descent optimizer to minimize the loss function and improve the accuracy of the prediction.

2. The automatic driving system for automobiles according to claim 1, characterized in that: The specific steps of the sensor and environment perception module are as follows: (1) Data processing: Sensors are used to collect information about the surrounding environment in real time, and the collected data is processed and analyzed to extract the location, speed, and direction information of road users; (2) Environmental model construction: Based on the processed data, an accurate environmental model is constructed to provide data support for the predictive driving algorithm.

3. The automatic driving system for automobiles according to claim 1, characterized in that: The specific steps of the decision-making planning and control module are as follows: (1) Risk assessment: Receive the predicted behavior of road users from the behavior prediction algorithm module and assess the potential conflicts and dangers between the autonomous vehicle and other road users based on the prediction results; (2) Issuing decision instructions: Based on the risk assessment results, a decision plan for the autonomous vehicle is generated, the decision plan is converted into specific control instructions, and the driving state of the autonomous vehicle is adjusted through the control execution module.

4. The automatic driving system for automobile according to claim 1, characterized in that: The specific steps of the learning and optimization module are as follows: (1) Data collection: Collect driving data of the autonomous vehicle, as well as feedback information from sensors and other modules to understand the performance of the autonomous vehicle in actual operation; (2) Model adjustment and optimization: Based on the collected data and feedback information, the predictive driving algorithm is adjusted and optimized to improve prediction accuracy and performance.

5. The automatic driving system for automobile according to claim 3, characterized in that The decision planning and control module solves the optimal decision planning trajectory through the dynamic programming algorithm, and receives the output of the behavior prediction algorithm module, that is, the prediction results of the future behavior of other road users. Let the prediction result set be P = {p1, p2, ..., p n }, where p i represents the predicted behavior trajectory of the i-th road user. The module evaluates potential conflicts and dangers by calculating the distance and speed difference between the autonomous vehicle and each predicted behavior trajectory. The calculation formula is: Among them, CarPos t represents the position of the autonomous vehicle at time t, p i (t) represents the predicted behavior trajectory p i At time t, T represents the time range of the evaluation. After evaluating potential conflicts and dangers, the decision planning and control module uses a dynamic programming-based algorithm to make decisions. Let the state space of the autonomous vehicle be X, the control input space be U, and the cost function be J(x i ,u ∈ ), where x i ∈X represents the current state, u ∈ ∈U represents the control input, and the calculation formula is: J(x i ,u ∈ )= is the weight coefficient. For each time step t, the module calculates t Starting from the cost function value under all possible control input sequences, the control input sequence that minimizes the cost function is selected as the optimal decision planning trajectory. By iteratively solving the optimal control input at each time step, the optimal decision planning trajectory is finally obtained.

6. The automatic driving system for automobile according to claim 4, characterized in that The learning and optimization module uses the gradient descent algorithm to update the model parameters by calculating the gradient of the loss function with respect to the model parameters and iteratively updating the parameters according to the learning rate. The calculation formula of the loss function is: Specifically, for each new sample (x i ,y i ), calculate the gradient of the loss function with respect to the model parameters, Update the parameters according to the learning rate η: θ By continuously iterating and updating parameters to gradually reduce prediction errors and improve prediction accuracy, the updated model will be deployed in autonomous vehicles for real-time prediction and decision-making. By working in collaboration with other modules, autonomous vehicles can better adapt to changes in the actual traffic environment and improve safety and reliability.

7. The automatic driving system for automobile according to claim 2, characterized in that The sensor and environment perception module uses the target detection algorithm of the YOLO framework to locate and classify the target. Let the input image be I and the feature vector extracted by the convolutional neural network be x. The target detection algorithm calculates the score and bounding box coordinates of the target through the following formula: Among them, S represents the score of the target, σ represents the activation function, which is used to map the score to the interval [0,1], and W s represents the weight of the scoring layer, x represents the feature vector, and B represents the coordinates and size of the bounding box, including the coordinates of the center point (x c ,y c ) width w and height h, W b Represents the weight of the bounding box layer. During the training process, W is continuously updated through the back propagation algorithm and gradient descent optimization method. s and W b The weight parameters enable the model to accurately identify the target in the image.

8. The automatic driving system for a vehicle according to claim 1, characterized in that ,The system also includes a human-computer interaction interface for displaying ,prediction results, decision planning information, and providing a user control and ,intervention interface, so that the user can understand the operating status and ,decision-making process of the autonomous vehicle and perform manual ,intervention when necessary.

Images