An intelligent traffic flow management system and method for unmanned vehicles

Abstract

The application belongs to the technical field of unmanned driving, and particularly relates to an intelligent traffic flow management system and method for unmanned vehicles. Firstly, environmental data and traffic flow data around the vehicle are collected and preprocessed. Then, obstacle detection and identification are performed to detect and identify characteristic indexes of the traffic flow data. The output result is used to predict the probability of the current vehicle passing through the obstacle and predict the future traffic flow dynamics. Then, an adaptive algorithm is called to calculate the dynamic passing threshold of the target vehicle according to the probability of the vehicle passing through the obstacle, the obstacle detection result and the traffic index identification result. Whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold is judged. If yes, a warning command is generated and transmitted to the background. If no, the dynamic behavior of the target vehicle is generated according to the predicted future traffic flow dynamics. The application can solve the problem that the evaluation of the influence of obstacles on the passing function of the unmanned driving technology in the prior art is not accurate.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of unmanned driving, and particularly relates to an intelligent traffic flow management system and method for unmanned vehicles. BACKGROUND

[0002] With the rapid development of artificial intelligence and automation technology, unmanned vehicles are gradually becoming an important part of modern transportation systems. Unmanned vehicles are those that can autonomously complete tasks such as navigation, driving, and parking without the direct operation of a human driver. They use advanced sensors such as lidar, cameras, radar, and ultrasonic sensors, GPS positioning, map data, and complex algorithms to perceive the surrounding environment and make decisions. There are already various types of unmanned vehicles on the market, including but not limited to unmanned delivery vehicles, unmanned courier vehicles, and unmanned taxis.

[0003] Despite the significant progress made in unmanned technology, these vehicles still face some challenges when faced with complex and variable real-world traffic conditions. First, the reaction of unmanned vehicles in dealing with certain specific obstacles may not be intelligent or flexible enough, for example:

[0004] 1. When a small obstacle (such as a bag) appears in front of the vehicle, the unmanned vehicle may take overly conservative action such as stopping immediately due to its inability to accurately assess the impact on traffic, which can cause unnecessary traffic disruption;

[0005] 2. When trying to make a turning maneuver (such as a left turn) and encountering a roadblock, the unmanned vehicle may choose to wait in place rather than finding an alternative route, causing local traffic congestion;

[0006] 3. When encountering a situation that requires detouring (such as a road being partially closed due to illegal parking), the unmanned vehicle may cause the entire road to be blocked due to excessive caution to accommodate other vehicles;

[0007] 4. In adverse weather conditions (such as heavy rain), the unmanned vehicle may have difficulty correctly judging the depth of the water and its impact on safe driving, and sometimes even risks proceeding, increasing potential risks.

[0008] The reason for the above problems is that existing unmanned technology does not accurately assess the impact of obstacles on traffic functionality, and lacks effective integration of real-time traffic information. SUMMARY

[0009] The technical problem solved by the present application is to provide an intelligent traffic flow management system and method for unmanned vehicles to address the problem of inaccurate assessment of the impact of obstacles on traffic functionality in existing technology.

[0010] The basic scheme provided by the present application is an intelligent traffic flow management system for unmanned vehicles, comprising:

[0011] a data acquisition module for acquiring environmental data around the vehicle and current traffic flow data of the vehicle;

[0012] a preprocessing module for preprocessing the acquired environmental data and traffic flow data to generate preprocessed environmental data and preprocessed traffic flow data, respectively;

[0013] a target detection module for obstacle detection and identification of the preprocessed environmental data according to a preset target detection algorithm, outputting an obstacle detection result, and feature index detection and identification of the preprocessed traffic flow data, outputting a traffic index identification result;

[0014] a passing prediction module for predicting the probability of the current vehicle passing through the obstacle using a prediction model based on the obstacle detection result, and predicting future traffic flow dynamics based on the traffic index identification result;

[0015] a passing threshold calculation module for calculating the dynamic passing threshold of the target vehicle based on the predicted probability of the current vehicle passing through the obstacle, the obstacle detection result, and the traffic index identification result using an adaptive algorithm;

[0016] a passing judgment module for judging whether the dynamic passing threshold of the target vehicle exceeds a preset passing threshold, and if so, generating a warning command; if not, generating a dynamic behavior of the target vehicle based on the predicted future traffic flow dynamics;

[0017] an execution control module for generating a control command based on the dynamic behavior of the target vehicle and transmitting the control command to the control system of the target vehicle, and transmitting the warning command to the target vehicle control background.

[0018] Further, the data acquisition module comprises a vehicle environmental data acquisition unit and a traffic monitoring unit, the vehicle environmental data acquisition unit is used to acquire environmental data around the vehicle through the sensor unit and the camera unit arranged on the vehicle body;

[0019] the traffic monitoring unit is used to generate traffic flow data of the current road of the vehicle by receiving real-time traffic data transmitted by the traffic monitoring device.

[0020] Further, the passing threshold calculation module comprises a threshold setting unit and a threshold adjustment unit, the threshold setting unit is used to set the passing threshold as a preliminary preset passing threshold according to the passing threshold reached during the training process of the prediction model in the passing prediction module;

[0021] The threshold adjustment unit is configured to dynamically adjust the preliminary preset passing threshold according to the predicted current vehicle passing obstacle probability, the obstacle detection result, and the traffic indicator recognition result, and generate a dynamic passing threshold.

[0022] Further, the threshold adjustment unit dynamically adjusts the preliminary preset passing threshold according to the predicted current vehicle passing obstacle probability, the obstacle detection result, and the traffic indicator recognition result, and generates a dynamic passing threshold, specifically as follows.

[0023] Obtain the current predicted current vehicle passing obstacle probability and the obstacle detection result, and assign a risk score to different types of obstacles.

[0024] Retrieve the environmental parameters in the target vehicle passing area traffic indicator recognition result, quantitatively process and weight distribute the environmental parameters, generate a target vehicle environmental complexity value, and map the generated target vehicle environmental complexity value to a preset risk score range to generate a target vehicle environmental complexity risk score.

[0025] Combine the obstacle risk score and the target vehicle environmental complexity risk score, and dynamically adjust the preliminary preset passing threshold according to a preset adaptive threshold algorithm.

[0026] Further, the obtaining of the current predicted current vehicle passing obstacle probability and the obstacle detection result, and the assignment of a risk score to different types of obstacles are specifically as follows.

[0027] Obtain the obstacle detection result and preset different types of obstacle risk level ranges.

[0028] Assign passing probability thresholds to different types of obstacles from the current vehicle passing obstacle probability.

[0029] Map the passing probability thresholds of different types of obstacles to the preset different types of obstacle risk level ranges to generate different types of obstacle risk scores, with the expression being:

[0030] Pobstacle = L min + [G × (L max - L min )]

[0031] Wherein, Pobstacle represents the obstacle risk score, L min represents the lowest risk level of the obstacle preset risk level range, L max represents the highest risk level of the obstacle preset risk level range, and G represents the obstacle passing probability threshold.

[0032] Furthermore, the step of retrieving environmental parameters from the traffic indicator identification results of the target vehicle's passage area, quantifying and weighting these parameters to generate an environmental complexity value for the target vehicle, and mapping this value to a preset risk scoring range to generate a target vehicle environmental complexity risk score, specifically involves:

[0033] The environmental parameters are obtained from the traffic indicator recognition results within the target vehicle's travel area. These environmental parameters include traffic flow, road conditions, and route complexity.

[0034] Traffic flow, road conditions, and route complexity are quantified, and corresponding weights are assigned to the quantified traffic flow, road conditions, and route complexity according to the preset parameter importance.

[0035] A preset environmental scoring range is defined. Weighted scores are then calculated for traffic flow, road conditions, and route complexity to generate an environmental complexity value. The calculation formula is as follows:

[0036] H ring = w1 × P car + w2 × D lane + w3 × L road

[0037] Wherein, H represents the environmental complexity value, P represents the traffic flow score, w1 represents the traffic flow parameter weight, D represents the road condition score, w2 represents the road condition parameter weight, L represents the route complexity score, and w3 represents the route complexity parameter weight.

[0038] A preset risk scoring range is defined, and environmental complexity values ​​are mapped to this range to generate an environmental complexity risk score for the target vehicle. The expression is as follows:

[0039] P_ring = (H_ring - H_ring) -min ) / (H ring -max -H ring -min )×(S max -S min )+S min

[0040] Wherein, P-ring represents the environmental complexity risk score, and H-ring represents the environmental complexity value. -min H represents the minimum value within the preset environmental scoring range. -max S represents the maximum value of the preset environmental score range. max S represents the maximum value of the preset risk score range. min This indicates the minimum value within the preset risk score range.

[0041] Furthermore, the expression for dynamically adjusting the initially preset passage threshold based on the preset adaptive threshold algorithm, by combining obstacle risk scores and target vehicle environmental complexity risk scores, is as follows:

[0042] Z = Z0+ w4xPbarriers + w5xPcircumstances

[0043] wherein, Z represents a dynamic passing threshold, Z0represents a preliminary preset passing threshold, w4represents a weight of a barrier risk score, and w5represents a weight of an environmental complexity risk score.

[0044] Further, the passing judgment module comprises a judgment unit, a pre-warning triggering unit, and a vehicle action adjustment unit, the judgment unit is configured to judge whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold, and output a judgment result;

[0045] The pre-warning triggering unit is configured to trigger a preset pre-warning instruction to generate a pre-warning command when the dynamic passing threshold of the target vehicle exceeds the preset passing threshold in the judgment result;

[0046] The vehicle action adjustment unit is configured to analyze and generate the dynamic behavior of the target vehicle according to the predicted future traffic flow when the dynamic passing threshold of the target vehicle does not exceed the preset passing threshold in the judgment result.

[0047] An intelligent traffic flow management method for unmanned vehicles, applied to the intelligent traffic flow management system for unmanned vehicles, comprising:

[0048] S1: collecting environmental data around the vehicle and current traffic flow data of the vehicle, and preprocessing to generate preprocessed environmental data and preprocessed traffic flow data;

[0049] S2: calling a preset target detection algorithm to detect and identify obstacles in the preprocessed environmental data, and to detect and identify feature indicators in the preprocessed traffic flow data, and outputting obstacle detection results and traffic indicator identification results respectively;

[0050] S3: calling a prediction model to predict the probability of the current vehicle passing through the obstacles according to the obstacle detection results, and to predict the future traffic flow dynamics according to the traffic indicator identification results;

[0051] S4: calling an adaptive algorithm to calculate the dynamic passing threshold of the target vehicle according to the predicted probability of the current vehicle passing through the obstacles, the obstacle detection results, and the traffic indicator identification results;

[0052] S5: judging whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold, if not, generating a pre-warning command; if yes, generating a dynamic behavior of the target vehicle according to the predicted future traffic flow dynamics;

[0053] S6: transmitting the control command of the dynamic behavior of the target vehicle to the control system of the target vehicle for execution, and transmitting the pre-warning command to the control background of the target vehicle.

[0054] The principles and advantages of the present application are that in the present application, the environmental data around the vehicle and the current traffic flow data are first collected, for the environmental data, whether there is an obstacle near the vehicle can be detected and identified, for the traffic flow data, the traffic flow dynamics near the current vehicle can be predicted; and for the identified obstacle, the passing probability is predicted through the prediction model, at this time, the preliminary passing probability is not sufficient to meet the passing risk assessment of the real environment where the vehicle is located, for this, an adaptive algorithm is used, the environmental parameters of the target vehicle, the obstacle detection result and the preliminary passing probability are comprehensively considered, a dynamic passing threshold is generated to meet the real passing threshold of the target vehicle under the influence of the external environment in the real running condition, when the dynamic passing threshold does not meet the preset passing threshold, the running state of the vehicle is controlled in combination with the traffic flow dynamics, the passing instruction of the vehicle in the future period of time is generated, and when the dynamic passing threshold is met, a warning command is generated and transmitted to the background, the staff located in the background can remotely control the vehicle according to the content in the transmitted warning command, so that the vehicle can better face the situation of being unable to pass through the obstacle by itself.

[0055] And in the above whole process, the dynamic adjustment of the passing threshold principle set by the present application is upward feedback, by combining the vehicle passing probability and the traffic flow dynamics prediction, different operation instructions in different states are realized, so that the intelligent passing function of the unmanned vehicle is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The functional block diagram of the embodiment of the present application is shown in the figure;

[0057] Figure 2 The flowchart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0058] The following will be further described in detail through specific embodiments:

[0059] With the development of intelligence, more and more unmanned vehicles appear on the market, such as unmanned delivery vehicles, unmanned express vehicles, unmanned taxis, etc., and in these unmanned vehicles, when facing complex and changeable traffic conditions, the actual influence of obstacles on passing cannot be accurately evaluated, for example, when a roadblock appears at a turning intersection, the unmanned vehicle often waits in place with double flashing lights; for example, when encountering a parked vehicle that needs to be bypassed, the road is congested because of courtesy to oncoming vehicles; there are many such conditions, the reason is that the unmanned vehicle is not accurate in judging the road obstacles, and lacks real-time information of regional traffic and road conditions, under the influence of these two factors, the unmanned vehicle intelligent passing is insufficient.

[0060] For this, the technical solution set by the present application is an intelligent traffic flow management system for unmanned vehicles, such asFigure 1 As shown, it is composed of a data acquisition module, a preprocessing module, a target detection module, a traffic prediction module, a traffic judgment module and an execution control module. The data acquisition module is used to acquire the environmental data around the vehicle and the current traffic flow data of the vehicle. Specifically, the data acquisition module is composed of a vehicle environmental data acquisition unit and a traffic monitoring unit. The vehicle environmental data unit acquires the environmental data around the vehicle through the sensor unit and the camera unit arranged on the vehicle body, such as laser radar, ultrasonic radar, millimeter wave radar, camera, etc. Through the combination of radar sensor unit and camera unit, and according to the live adjustment of the installation position and angle of each radar sensor unit and camera unit, the image data and radar data on the road are effectively captured to obtain the environmental data around the target vehicle.

[0061] The traffic monitoring unit generates the traffic flow data of the current road of the vehicle by receiving the real-time traffic data transmitted by the traffic monitoring device. In this embodiment, the traffic monitoring unit receives the real-time traffic data such as traffic flow, vehicle speed, road condition, etc. transmitted back by the traffic monitoring camera, sensor and other devices through wireless transmission, so as to obtain the traffic flow data in real time.

[0062] The preprocessing module is used to preprocess the collected environmental data and traffic flow data to generate preprocessed environmental data and preprocessed traffic flow data, respectively. In this embodiment, the environmental data is mainly preprocessed image data, including color space conversion, converting RGB image to grayscale image; it also includes using Gaussian filter to remove noise in the image, and using histogram equalization to improve the contrast of the image.

[0063] The preprocessing of the traffic flow data is data cleaning, data supplementing and data conversion of the collected original real-time traffic data, so as to ensure the accuracy and integrity of the traffic flow data.

[0064] The preprocessed environmental data and traffic flow data are transmitted to the target detection module, and the target detection module detects and identifies the obstacles in the preprocessed environmental data according to the preset target detection algorithm, outputs the obstacle detection result, and detects and identifies the feature index of the preprocessed traffic flow data, outputs the traffic index identification result. Specifically, the target detection algorithm used for obstacle detection and identification of preprocessed environmental data is FasterR-CNN model, and its operation principle is as follows:

[0065] First, collect an image data set containing different obstacles. The collected image data set about obstacles should cover various scenes and conditions, such as obstacles under different weather and light conditions, and label the collected images, mark the position and category of obstacles in each image;

[0066] Then the image dataset is preprocessed, including size adjustment, cropping, color enhancement, standardization, etc., and data augmentation techniques are used, including random cropping, rotation, flipping, color transformation, etc., to generate more diverse image training datasets, which can help improve the robustness and generalization ability of the Faster R-CNN model in the target detection algorithm;

[0067] Subsequently, the Faster R-CNN model initializes the weights, and the generated diverse image training dataset is input into the Faster R-CNN model for training. During the training process, the weights are continuously adjusted through the backpropagation algorithm to optimize the loss function, so as to improve the recognition and classification accuracy of obstacles, and the learning rate, batch size, iteration number, etc. Hyperparameters are adjusted to optimize the training effect of the model;

[0068] Finally, the reserved validation set is input into the trained model to evaluate the performance of the model, including accuracy, recall rate, F1-score, etc. The performance of the model on the validation set is analyzed to identify and solve possible problems in the model. According to the evaluation results and feedback from error analysis, the model architecture, data processing process or training strategy is adjusted to further improve the performance of the model.

[0069] For feature index detection and identification of preprocessed traffic flow data, the present application uses a CNN-RNN hybrid model to analyze and model the preprocessed traffic flow data to obtain current traffic flow indicators and road condition indicators. The obtained current feature indicators are helpful for subsequent prediction of traffic flow dynamics.

[0070] The passing prediction module uses a prediction model to predict the probability of the vehicle passing through the obstacle based on the obstacle detection result. Specifically, the vehicle passing obstacle probability prediction process is as follows:

[0071] First, collect actual data of vehicles passing through obstacles, including vehicle speed, type of obstacles (such as pedestrians, vehicles, roadblocks, etc.), position, size, shape, etc. Information of obstacles;

[0072] Then, the corresponding features are extracted from the collected data, such as distance features (distance between vehicle and obstacle), speed features (vehicle speed and acceleration), obstacle attribute features (type, size, shape, etc. of obstacles), environmental features (weather conditions, road conditions, etc.); and the feature data is preprocessed, including missing value processing, outlier processing, converting non-numeric features into numeric features using one-hot encoding or label encoding, and finally generating training set and test set;

[0073] Then a prediction model such as a linear regression model, a logistic regression model or a random forest model is selected, and the training set is input into the model for training. During the training process, the model parameters are optimized. After the training is completed, the test set is used for testing to evaluate whether the indicators of the model meet the standards, such as mean squared error, mean absolute error, accuracy, recall rate, F1-score, etc.

[0074] Finally, according to the training results of different models, the best-performing model is selected to predict the obstacle detection results collected by the vehicle in real time, and the probability of the vehicle passing through the obstacle is obtained.

[0075] Meanwhile, the traffic prediction module also predicts the future traffic flow dynamics according to the traffic indicator recognition results. In this application, a CNN-RNN hybrid model is used to predict the future traffic flow dynamics, such as road congestion, smoothness, etc. The CNN-RNN hybrid model used in this application uses the CNN part to process the image or video data transmitted back by the traffic monitoring device and extract high-level features from it. The RNN part receives the feature representation from the CNN or processes time series data, and predicts the future traffic state according to these features and historical traffic data.

[0076] For the above-mentioned vehicle obstacle passing probability result, the traffic threshold calculation module calls an adaptive algorithm to calculate the dynamic traffic threshold of the target vehicle according to the predicted current vehicle obstacle passing probability, obstacle detection result and traffic indicator recognition result. Specifically, the traffic threshold calculation module includes a threshold setting unit and a threshold adjustment unit. The threshold setting unit is used to set the traffic threshold that meets the preset requirements during the training process of the prediction model in the traffic prediction module as the preliminary preset traffic threshold. For example, a threshold is selected to balance the accuracy and recall rate of the prediction model, and the threshold is tested in a simulated environment to evaluate its performance in the simulated environment. The threshold with the least false positives or false negatives is considered the best, and the preliminary preset traffic threshold is obtained.

[0077] However, the preliminary preset traffic threshold is not suitable for independent target vehicles in actual use. Therefore, the threshold adjustment unit adjusts the preliminary preset traffic threshold according to the predicted current vehicle obstacle passing probability, obstacle detection result and traffic indicator recognition result to generate a dynamic traffic threshold. In this embodiment, the threshold adjustment unit implements the following steps:

[0078] Step 1: Obtain the current predicted probability of the vehicle passing through the obstacle and the obstacle detection results, and assign risk scores to different types of obstacles. Specifically, first, obtain the obstacle detection results and preset the risk level range for different types of obstacles; for example, the risk level range for obstacle A is set to 1-5, and the risk level range for obstacle B is set to 6-10. Next, assign a passage probability threshold to different types of obstacles from the current vehicle passage probability; for example, a passage probability of obstacle A below 0.3 is considered high risk, 0.3-0.7 is considered medium risk, and above 0.7 is considered low risk. Finally, map the passage probability thresholds for different types of obstacles to the preset risk level ranges for different types of obstacles to generate risk scores for different types of obstacles, expressed as:

[0079] P barrier = L min +[G×(L max -L min )]

[0080] Where P represents the obstacle risk score, L... min L represents the lowest risk level within the preset risk level range for obstacles. max This indicates the highest risk level within the preset risk level range for obstacles, where G represents the obstacle passage probability threshold.

[0081] For example, if the probability of passage for obstacle A is 0.4 and the risk level ranges from 1 to 5, then:

[0082] P barrier -A =1 + [0.4*(5-1)] = 2.6

[0083] Step Two: Retrieve environmental parameters from the traffic indicator identification results of the target vehicle's passage area, quantify and weight these parameters to generate an environmental complexity value for the target vehicle, and map this value to a preset risk scoring range to generate an environmental complexity risk score for the target vehicle; specifically:

[0084] The environmental parameters are obtained from the traffic indicator recognition results within the target vehicle's travel area. These environmental parameters include traffic flow, road conditions (congestion level, road smoothness), and route complexity (number of intersections, turning angle).

[0085] Traffic flow, road conditions, and route complexity are quantified, and corresponding weights are assigned to the quantified traffic flow, road conditions, and route complexity according to preset parameter importance. Among them, traffic flow is quantified as the number of vehicles passing through a certain road segment per hour, road conditions are assessed by road maintenance status and weather, and route complexity is assessed by the number of intersections and the tortuosity of the road. Then, weights are assigned according to the importance of each parameter. In this embodiment, the weights of the three indicators are summed to 1.

[0086] The environmental scoring range is preset. In this embodiment, the scoring range of environmental parameters is consistent. For example, in the traffic flow scoring range, the levels are divided based on the number of vehicles per hour: less than 100 vehicles is 1 point, 100-500 is 5 points, and more than 500 is 10 points. In the road condition scoring range, the scoring is based on the road surface smoothness and maintenance status: excellent road is 1 point, ordinary is 5 points, and poor is 10 points. In the route complexity scoring range, the scoring is based on the number of intersections and the degree of curvature: few intersections and a straight path is 1 point, and many intersections and a curvature are 10 points.

[0087] Then, the weighted scores for traffic flow, road conditions, and route complexity are calculated to generate an environmental complexity value. The calculation formula is as follows:

[0088] H ring = w1 × P car + w2 × D lane + w3 × L road

[0089] Wherein, H represents the environmental complexity value, P represents the traffic flow score, w1 represents the traffic flow parameter weight, D represents the road condition score, w2 represents the road condition parameter weight, L represents the route complexity score, and w3 represents the route complexity parameter weight.

[0090] A preset risk scoring range is defined in this embodiment, ranging from 1 to 10, where 1 represents the lowest risk and 10 represents the highest risk. Environmental complexity values ​​are mapped to this preset risk scoring range to generate a target vehicle environmental complexity risk score, expressed as:

[0091] P_ring = (H_ring - H_ring) -min ) / (H ring -max -H ring -min )×(S max -S min )+S min

[0092] Wherein, P-ring represents the environmental complexity risk score, and H-ring represents the environmental complexity value. -min H represents the minimum value within the preset environmental scoring range. -max S represents the maximum value of the preset environmental score range. max S represents the maximum value of the preset risk score range. min This indicates the minimum value within the preset risk score range.

[0093] For example, the minimum value of environmental complexity is 0, and the maximum value is 100; the minimum value of risk score is 1, and the maximum value is 10. If the calculated environmental complexity is 75, then:

[0094]

[0095] The preliminary preset passing threshold is dynamically adjusted according to a preset adaptive threshold algorithm in combination with the obstacle risk score and the target vehicle environment complexity risk score, and the expression is:

[0096] Z = Z0 + w4 * Pobst + w5 * Penv

[0097] wherein Z represents the dynamic passing threshold, Z0 represents the preliminary preset passing threshold, w4 represents the weight of the obstacle risk score, and w5 represents the weight of the environment complexity risk score.

[0098] After the dynamic passing threshold of the target vehicle is calculated, the passing judgment module is used to judge whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold, and if yes, a warning command is generated; if no, the dynamic behavior of the target vehicle is dynamically generated according to the predicted future traffic flow; wherein the passing judgment module comprises a judgment unit, a warning triggering unit and a vehicle action adjustment unit, the judgment unit is used to judge whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold, and output the judgment result;

[0099] The warning triggering unit is used to trigger the preset warning instruction to generate a warning command when the dynamic passing threshold of the target vehicle exceeds the preset passing threshold in the judgment result;

[0100] The vehicle action adjustment unit is used to retrieve the predicted future traffic flow dynamics to analyze and generate the dynamic behavior of the target vehicle when the dynamic passing threshold of the target vehicle does not exceed the preset passing threshold in the judgment result. In this embodiment, according to different ranges of the dynamic passing threshold, the warning levels are set, respectively including:

[0101] Normal (low risk): the dynamic passing threshold is lower than the preset passing threshold, for example, 10;

[0102] Warning (medium risk): the dynamic passing threshold is between two preset passing thresholds, for example, 10-15;

[0103] Danger (high risk): the dynamic passing threshold is higher than the preset passing threshold, for example, 15;

[0104] Then, it is determined that the specific dynamic passing threshold is located in which warning level, for example, the preset initial threshold is 5, the obstacle risk score calculation result is 6, the environment complexity risk score is 4, the weight coefficient w4 is 0.4, and w5 is 0.6, then the calculated adaptive threshold is:

[0105] Z = 5 + 0.4 * 6 + 0.6 * 4 = 9.8

[0106] Therefore, the calculated dynamic passing threshold is 9.8, and the warning level is normal; specifically, in each warning level, when:

[0107] If the dynamic passing threshold is less than or equal to 10, the target vehicle system continues to operate normally, and the relevant information of "normal state" or "no need for early warning" is displayed; at this time, the driving path and driving state of the target vehicle are generated according to the predicted future traffic flow dynamics;

[0108] If the dynamic passing threshold is greater than 10 and less than 15, a departure warning is given, and the information of "warning: risk increased" is displayed, and a corresponding warning early warning command is generated; under this early warning command, the background personnel make a decision whether to need manual intervention;

[0109] If the dynamic passing threshold is greater than 15, a high-risk departure warning is given, and the information of "danger: need to take action manually" is displayed, and a corresponding early warning command is generated.

[0110] The execution control module generates a control command according to the dynamic behavior of the target vehicle and transmits it to the control system of the target vehicle, and transmits the early warning command to the target vehicle control background, and makes different decisions under different early warning commands, so as to achieve the purpose of dynamically adjusting the running state of the target vehicle.

[0111] As shown in Figure 2 Another embodiment of the present embodiment also includes an intelligent traffic flow management method for unmanned vehicles, applied to the intelligent traffic flow management system for unmanned vehicles described above, comprising:

[0112] S1: Collecting the environmental data around the vehicle and the current traffic flow data of the vehicle, and preprocessing to generate preprocessed environmental data and preprocessed traffic flow data;

[0113] S2: Calling a preset target detection algorithm to detect and identify obstacles in the preprocessed environmental data, and to detect and identify feature indicators in the preprocessed traffic flow data, and outputting the obstacle detection result and the traffic indicator identification result respectively;

[0114] S3: Calling a prediction model to predict the probability of the current vehicle passing through the obstacle based on the obstacle detection result, and to predict the future traffic flow dynamics based on the traffic indicator identification result;

[0115] S4: Calling an adaptive algorithm to calculate the dynamic passing threshold of the target vehicle according to the predicted probability of the current vehicle passing through the obstacle, the obstacle detection result, and the traffic indicator identification result;

[0116] S5: Determining whether the dynamic passing threshold of the target vehicle exceeds the preset passing threshold, if yes, generating an early warning command; if not, generating the dynamic behavior of the target vehicle according to the predicted future traffic flow dynamics;

[0117] S6: transmitting the dynamic behavior generation control command of the target vehicle to the control system of the target vehicle for execution, and transmitting the early warning command to the target vehicle control background.

[0118] The above is only an embodiment of the present application, and the common knowledge of specific structures and characteristics in the scheme is not described in detail. The ordinary skilled person in the art knows all the ordinary technical knowledge in the field of the present application before the filing date or the priority date, can know all the prior art in the field, and has the ability to apply conventional experimental means before that date. The ordinary skilled person in the art can perfect and implement the present scheme based on their own ability under the guidance of the present application. Some typical known structures or known methods should not be an obstacle for the ordinary skilled person in the art to implement the present application. It should be noted that, for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should also be considered as the protection scope of the present application. The scope of protection claimed in the present application should be subject to the content of its claims, and the specific implementation in the specification can be used to explain the content of the claims.

Claims

1. An intelligent traffic flow management system for autonomous vehicles, characterized in that: include: The data acquisition module is used to collect environmental data around the vehicle and the vehicle's current traffic flow data; The preprocessing module is used to preprocess the collected environmental data and traffic flow data to generate preprocessed environmental data and preprocessed traffic flow data, respectively. The target detection module is used to detect and identify obstacles in the preprocessed environmental data according to a preset target detection algorithm and output the obstacle detection results. In addition, feature index detection and identification are performed on the preprocessed traffic flow data, and traffic index identification results are output. The traffic prediction module is used to predict the probability of a vehicle passing an obstacle based on the obstacle detection results using a prediction model, and to predict future traffic flow dynamics based on traffic indicator identification results. The passage threshold calculation module calls an adaptive algorithm to calculate the dynamic passage threshold of the target vehicle based on the predicted probability of the current vehicle passing through the obstacle, the obstacle detection result, and the traffic indicator recognition result. The passage judgment module is used to determine whether the dynamic passage threshold of the target vehicle exceeds the preset passage threshold. If so, it generates a warning command. If not, the dynamic behavior of the target vehicle is generated based on the predicted future traffic flow. The execution control module is used to generate control commands based on the dynamic behavior of the target vehicle and transmit them to the target vehicle's control system, as well as to transmit warning commands to the target vehicle's control backend. The passage threshold calculation module includes a threshold setting unit and a threshold adjustment unit. The threshold setting unit is used to set the passage threshold that meets the preset requirements during the training of the prediction model in the passage prediction module as the initial preset passage threshold. The threshold adjustment unit is used to dynamically adjust the initially preset passage threshold based on the predicted probability of the current vehicle passing through the obstacle, the obstacle detection result, and the traffic indicator recognition result, thereby generating a dynamic passage threshold. The threshold adjustment unit dynamically adjusts the initially preset passage threshold based on the predicted probability of the current vehicle passing the obstacle, the obstacle detection result, and the traffic indicator recognition result, generating a dynamic passage threshold as follows: Obtain the current predicted probability of the vehicle passing through the obstacle and the obstacle detection results, and assign risk scores to different types of obstacles; The environmental parameters in the traffic indicator identification results of the target vehicle's passage area are retrieved, and the environmental parameters are quantified and weighted to generate the environmental complexity value of the target vehicle. The generated environmental complexity value of the target vehicle is then mapped to a preset risk scoring range to generate the environmental complexity risk score of the target vehicle. By combining obstacle risk scores and target vehicle environmental complexity risk scores, the initially preset passage threshold is dynamically adjusted according to a preset adaptive threshold algorithm.

2. The intelligent traffic flow management system for unmanned vehicles according to claim 1, characterized in that: The data acquisition module includes a vehicle environment data acquisition unit and a traffic monitoring unit. The vehicle environment data acquisition unit is used to collect environmental data around the vehicle through sensor units and camera units deployed on the vehicle body. The traffic monitoring unit is used to generate traffic flow data of the current road by receiving real-time traffic data transmitted by traffic monitoring equipment.

3. The intelligent traffic flow management system for unmanned vehicles according to claim 2, characterized in that: The specific steps for obtaining the currently predicted probability of the vehicle passing through the obstacle, the obstacle detection results, and assigning risk scores to different types of obstacles are as follows: Obtain obstacle detection results and preset risk level ranges for different types of obstacles; Assign passage probability thresholds to different types of obstacles based on the current vehicle passage probability; The passage probability thresholds for different types of obstacles are mapped to preset risk level ranges for different types of obstacles to generate risk scores for different types of obstacles. The expression is as follows: in, Indicates obstacle risk score, This indicates the lowest risk level within the preset risk level range for the obstacle. This indicates the highest risk level within the preset risk level range for the obstacle. This represents the threshold for the probability of passing through an obstacle.

4. The intelligent traffic flow management system for unmanned vehicles according to claim 3, characterized in that: The process of retrieving environmental parameters from the traffic indicator identification results of the target vehicle's passage area, quantifying and weighting these parameters to generate an environmental complexity value for the target vehicle, and mapping this value to a preset risk scoring range to generate a target vehicle environmental complexity risk score is as follows: The environmental parameters are obtained from the traffic indicator recognition results within the target vehicle's travel area. These environmental parameters include traffic flow, road conditions, and route complexity. Traffic flow, road conditions, and route complexity are quantified, and corresponding weights are assigned to the quantified traffic flow, road conditions, and route complexity according to the preset parameter importance. A preset environmental scoring range is defined. Weighted scores are then calculated for traffic flow, road conditions, and route complexity to generate an environmental complexity value. The calculation formula is as follows: in, This represents the environmental complexity value. Indicates traffic flow score, Indicates the weight of the traffic flow parameter. Indicates road condition score, Indicates the weights of road condition parameters. Indicates the route complexity score. Indicates the weight of the route complexity parameter; A preset risk scoring range is defined, and environmental complexity values ​​are mapped to this range to generate an environmental complexity risk score for the target vehicle. The expression is as follows: in, This indicates the risk score for environmental complexity. This represents the environmental complexity value. This represents the minimum value within the preset environmental score range. This represents the maximum value within the preset environmental score range. This indicates the maximum value of the preset risk score range. This indicates the minimum value within the preset risk score range.

5. The intelligent traffic flow management system for unmanned vehicles according to claim 4, characterized in that: The expression for dynamically adjusting the initially preset passage threshold based on the combination of obstacle risk score and target vehicle environmental complexity risk score, according to the preset adaptive threshold algorithm, is as follows: in, Indicates the dynamic access threshold. This indicates the initially preset passage threshold. This indicates the weight of the obstacle risk score. This indicates the weight of the environmental complexity risk score.

6. The intelligent traffic flow management system for unmanned vehicles according to claim 5, characterized in that: The passage judgment module includes a judgment unit, a warning triggering unit, and a vehicle action adjustment unit. The judgment unit is used to judge whether the dynamic passage threshold of the target vehicle exceeds the preset passage threshold and output the judgment result. The warning triggering unit is used to trigger a preset warning command to generate a warning command when the dynamic passage threshold of the target vehicle exceeds the preset passage threshold in the judgment result. The vehicle motion adjustment unit is used to retrieve the predicted future traffic flow dynamics and analyze and generate the dynamic behavior of the target vehicle when the dynamic passage threshold of the target vehicle does not exceed the preset passage threshold in the judgment result.

7. An intelligent traffic flow management method for autonomous vehicles, applied to an intelligent traffic flow management system for autonomous vehicles as described in any one of claims 1-6, characterized in that: include: S1: Collect environmental data around the vehicle and the current traffic flow data of the vehicle, and preprocess the data to generate preprocessed environmental data and preprocessed traffic flow data. S2: Call the preset target detection algorithm to perform obstacle detection and recognition on the preprocessed environmental data, and perform feature index detection and recognition on the preprocessed traffic flow data, and output the obstacle detection results and traffic index recognition results respectively; S3: Call the prediction model to predict the probability of the current vehicle passing through the obstacle based on the obstacle detection results, and predict the future traffic flow dynamics based on the traffic indicator recognition results; S4: Call the adaptive algorithm to calculate the dynamic passage threshold of the target vehicle based on the predicted probability of the current vehicle passing through the obstacle, the obstacle detection result, and the traffic indicator recognition result; S5: Determine whether the dynamic passage threshold of the target vehicle exceeds the preset passage threshold. If so, generate a warning command. If not, the dynamic behavior of the target vehicle is generated based on the predicted future traffic flow. S6: Transmit the dynamic behavior generation control command of the target vehicle to the target vehicle's control system for execution, and transmit the warning command to the target vehicle's control backend.

Images