Vehicle dynamic driving scene judgment method based on multi-source information fusion

Abstract

The invention discloses a vehicle dynamic driving scene determination method based on multi-source information fusion, and the method comprises the following steps: obtaining a target driving path of a target vehicle, and dividing the target driving path into a plurality of continuous sub-regions; acquiring a multi-source information parameter of each sub-region, and outputting an estimated environment scene and a traffic state in each sub-region through a multi-modal combined neural network; based on the pre-judged environment scene and traffic state, calculating the probability that the vehicle has different travel working conditions in each sub-region; according to the scheme, dynamic balance of model complexity and prediction precision under segmented scenes is achieved, meanwhile, segmented probabilistic judgment results including congestion, start and stop and other working condition types are further output, a scene feature sequence along a driving path is formed, and the dynamic driving scene description including the environment scene, the traffic state and the working condition probability is obtained. And forming a dynamic description system containing multi-dimensional scene elements.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy vehicles, and in particular to a method for determining vehicle dynamic driving scenarios based on multi-source information fusion. Background Art

[0002] Electric vehicle remaining range prediction, a core technology for alleviating user range anxiety and supporting optimized charging load scheduling, currently faces a key challenge: the impact of dynamic driving scenarios on vehicle energy consumption during future journeys is highly uncertain. Existing single-point estimation algorithms calculate the expected remaining range based solely on historical energy consumption data or single sensor information, completely ignoring the real-time impact of dynamic driving scenarios that the vehicle may encounter in the future. For example, when a vehicle is about to enter a congested road section or encounter rainy weather, existing technologies are unable to predict the energy consumption changes in these scenarios based on real-time traffic flow or weather data, resulting in significant deviations between the predicted remaining range and the actual value. More importantly, existing solutions are unable to dynamically predict the environmental scenario, traffic conditions, and operating condition probabilities for each sub-region of the vehicle's future travel path by fusing multi-source information (such as high-precision maps, real-time traffic flow, weather data, and historical travel history). This severely limits the potential application of electric vehicles in intelligent transportation systems. Therefore, a method that can predict vehicle dynamic driving scenarios in segments is urgently needed. Summary of the Invention

[0003] In view of the above-mentioned prior art, the present invention provides a vehicle dynamic driving scene determination method based on multi-source information fusion, which mainly solves the technical problems existing in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, the technical solution of the embodiment of the present invention is implemented as follows: A method for determining a vehicle dynamic driving scene based on multi-source information fusion, the method comprising the following steps: Obtaining a target driving path of a target vehicle, and dividing the target driving path into a plurality of continuous sub-areas; Obtain multi-source information parameters for each sub-area and output the estimated environmental scene and traffic status in each sub-area through a multimodal combined neural network; Based on the predicted environmental scenarios and traffic conditions, the probability of vehicles experiencing different travel conditions in each sub-area is calculated; The output includes a dynamic driving scenario description of the environmental scenario, traffic status and operating condition probability.

[0005] Optionally, obtaining the target driving path of the target vehicle includes: obtaining the target trip of the target vehicle; inputting the target trip of the target vehicle into a preset online map path planning model to generate the target driving path of the target vehicle.

[0006] Optionally, dividing the target driving path into a plurality of sub-areas specifically includes: presetting a segmentation threshold, and dividing the target driving path into a plurality of sub-areas according to the segmentation threshold.

[0007] Optionally, the multi-source information parameters include high-precision map information, real-time traffic flow information, real-time weather information, and vehicle historical travel information.

[0008] Optionally, the multimodal combined neural network includes a scene attribute feature extraction network, a traffic behavior time series feature extraction network, and an environment space distribution feature extraction network. The scene attribute feature vector of the target sub-area is extracted by the scene attribute feature extraction network, the traffic behavior time series feature vector of the target sub-area is extracted by the traffic behavior time series feature extraction network, and the environment space distribution feature vector of the target sub-area is extracted by the environment space distribution feature extraction network.

[0009] Optionally, the scene attribute feature vector and the environment space distribution feature vector are input into the fully connected layer of the multimodal combination neural network for feature fusion. After fusion, the environment scene type and probability distribution are output through the Softmax classifier.

[0010] Optionally, a logistic regression model is established, and specific scene attribute feature vectors, traffic behavior time series feature vectors, and environmental spatial distribution feature vectors are selected to input into the logistic regression model, and the historical dominant probability and scene constraint probability are output. If the historical dominant probability is less than the scene constraint probability, the scene attribute feature vector and the traffic behavior time series feature vector are input into the fully connected layer in the multimodal combination neural network for feature fusion. After fusion, the traffic state type and probability distribution are output through the Softmax classifier.

[0011] Optionally, if the historical dominant probability is greater than the scenario constraint probability, the traffic behavior time series feature vector is input into the fully connected layer in the multimodal combination neural network, and the traffic state type and probability distribution are output.

[0012] Optionally, based on the predicted environmental scenarios and traffic conditions, the probability of different travel conditions for vehicles in each sub-area is calculated, specifically including: establishing a random forest model, inputting the environmental scenario type and probability distribution as well as the traffic condition type and probability distribution into the random forest model, and outputting the travel condition type and probability distribution.

[0013] The beneficial effects of the present invention are as follows: by acquiring the target driving path and dividing it into sub-areas, high-precision map information, real-time traffic flow, weather data and historical travel characteristics are simultaneously collected for each sub-area, and through the collaborative operation of the scene attributes, traffic behavior time series and environmental spatial distribution feature extraction network in the multimodal combination neural network, the distributed feature capture of the static attributes of the roads, traffic dynamic patterns and environmental heterogeneity of each sub-area is achieved, avoiding the lack of representation of local scene mutations in global modeling; through the mutual cooperation of the logistic regression model and the random forest model, while achieving a dynamic balance between the model complexity and prediction accuracy in the segmented scenario, it further outputs the segmented probabilistic judgment results including congestion, start-stop and other working conditions, forming a scene feature sequence along the driving path, and forming a dynamic description system including multi-dimensional scene elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a schematic diagram of the structure of the motor in the embodiment of the present application; DETAILED DESCRIPTION

[0015] The technical solution of the present invention is further elaborated in detail below in conjunction with the drawings and specific embodiments of the specification. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. In the following description, reference is made to "some embodiments", which describes a subset of all possible embodiments, but it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0016] In the following description, numerous specific details are provided to provide a more thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without one or more of these details. In other instances, certain technical features well known in the art are not described to avoid confusion with the present invention.

[0017] It should be understood that the present invention can be implemented in different forms and should not be interpreted as being limited to the embodiments proposed herein. On the contrary, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the present invention to those skilled in the art. And the purpose of the terms used herein is only to describe specific embodiments and is not intended to limit the present invention. When used herein, the singular forms "one", "an" and "said / the" are also intended to include plural forms, unless the context clearly indicates another way. It should also be understood that the terms "comprising" and / or "comprising" when used in this specification determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0018] It should also be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only implementation methods.

[0019] In order to fully understand the present invention, a detailed structure will be provided in the following description to illustrate the technical solution proposed by the present invention. Optional embodiments of the present invention are described in detail below. However, in addition to these detailed descriptions, the present invention may also have other implementations.

[0020] Example 1 Please refer to the attached Figure 1 The present application provides a method for determining a vehicle dynamic driving scene based on multi-source information fusion, the method comprising the following steps: S1. Obtain a target driving path of a target vehicle, and divide the target driving path into a plurality of continuous sub-areas; S2. Obtain multi-source information parameters for each sub-area and output the estimated environmental scene and traffic status in each sub-area through a multi-modal combined neural network; S3. Calculate the probability of different vehicle travel conditions in each sub-area based on the predicted environmental scenario and traffic conditions; S4. Output a dynamic driving scenario description including environmental scenarios, traffic conditions, and operating condition probabilities.

[0021] Specifically, after obtaining the target trip of the target vehicle, it is input into the preset online map path planning model to generate the target driving path. The path is divided into multiple continuous sub-areas based on the preset segmentation threshold. Multi-source parameters such as high-precision map information, real-time traffic flow information, real-time weather information and vehicle historical trip information are collected for each sub-area. With the help of the scene attribute feature extraction network, traffic behavior time series feature extraction network and environmental spatial distribution feature extraction network in the multimodal combination neural network, the corresponding feature vectors are extracted respectively; the scene attribute feature vector and the environmental spatial distribution feature vector are input into the fully connected layer for fusion, and the environmental scene type and probability distribution are output through the Softmax classifier. At the same time, a logistic regression model is established, and a specific feature vector is input to obtain the historical dominant probability and scene constraint probability. If the historical dominant probability is less than the scene constraint probability, the scene attribute feature vector and the traffic behavior time series feature vector are input into the fully connected layer for fusion, and then the traffic state type and probability distribution are output through the Softmax classifier; if the historical dominant probability is greater than the scene constraint probability, the traffic behavior time series feature vector is directly input into the fully connected layer and the corresponding traffic state type and probability distribution are output. Based on the predicted environmental scenarios and traffic conditions, a random forest model is constructed and the relevant types and probability distributions are input. The trip condition types and probability distributions are output. Finally, the environmental scenarios, traffic conditions, and condition probability information are integrated to form a dynamic driving scenario description.

[0022] In some embodiments, obtaining the target driving path of the target vehicle includes: obtaining a target range of the target vehicle; and inputting the target range of the target vehicle into a preset online map path planning model to generate the target driving path of the target vehicle.

[0023] Furthermore, the target driving path is divided into a plurality of sub-areas, specifically comprising: presetting a segmentation threshold, and dividing the target driving path into the plurality of sub-areas according to the segmentation threshold. For example, using 5 km as the segmentation threshold, the target driving path is divided into the plurality of sub-areas according to 5 km.

[0024] In some embodiments, the multi-source information parameters include high-precision map information, real-time traffic flow information, real-time weather information, and vehicle historical travel information.

[0025] High-precision map information encompasses precise spatial data such as road geometry, lane attributes, traffic signs and markings, and road infrastructure. It provides detailed information such as road curvature, slope, number of lanes, and traffic light locations and types, accurately describing the road's physical environment and traffic regulations. Real-time traffic flow information includes dynamic data such as traffic density, vehicle speed, congestion, and the location of accidents or construction sites on the current road section. It reflects road capacity and the real-time traffic status of vehicles in motion. Real-time weather information encompasses meteorological factors such as temperature, humidity, wind speed, precipitation, and visibility, clarifying the impact of current weather conditions on vehicle operation, such as the slipperiness of rainy roads and reduced visibility in foggy conditions. Historical vehicle travel information includes historical data such as the vehicle's past routes, speed changes, acceleration and deceleration frequency, energy consumption, and fault records. This information can be used to analyze the vehicle's driving patterns and performance in different scenarios, providing a reference for determining the current scenario.

[0026] In some embodiments, the multimodal combined neural network includes a scene attribute feature extraction network, a traffic behavior time series feature extraction network, and an environment space distribution feature extraction network. The scene attribute feature vector of the target sub-area is extracted by the scene attribute feature extraction network, the traffic behavior time series feature vector of the target sub-area is extracted by the traffic behavior time series feature extraction network, and the environment space distribution feature vector of the target sub-area is extracted by the environment space distribution feature extraction network.

[0027] Specifically, a multimodal combined neural network integrates heterogeneous data from multiple sources to construct a multidimensional perception architecture covering static environments, dynamic traffic, and spatial weather. The scene attribute feature extraction network uses a multi-layer perceptron to process high-precision map parameters and historical travel data. The input layer receives HD map information and vehicle historical travel information, including road geometry such as slope gradient, curvature radius, and lane width; road surface physical properties such as material friction coefficient and water depth; traffic infrastructure information such as signal phase and guardrail type; and historical vehicle behavior data such as accident frequency and sudden acceleration events. Key features are extracted using nonlinear activation functions such as Reluctant Unit (ReLU) in the hidden layer. For example, the network combines the number of lanes and slope to calculate road capacity characteristics, quantifies the multiplicative effect of curve curvature and visibility to form risk-sensitive features, and constructs driving behavior features by statistically analyzing the probability of sudden braking events on specific road sections. The network ultimately outputs a scene attribute feature vector representing the static geometric attributes and historical behavior characteristics of the sub-region.

[0028] The traffic behavior temporal feature extraction network receives real-time traffic flow information and processes the real-time traffic flow data by using a long short-term memory network. Based on dynamic traffic parameters such as vehicle speed mean, acceleration standard deviation, and headway distribution in the real-time traffic flow data, it filters out noise data such as abnormal sudden braking events through a forget gate, uses cell states to retain periodic long-term trends such as morning and evening rush hours, marks sudden congestion events as key timestamps through an output gate, captures temporal dependencies with the help of a gating mechanism, and outputs a traffic behavior temporal feature vector that reflects the dynamic characteristics of the traffic state evolving over time.

[0029] The environmental spatial distribution feature extraction network uses a convolutional neural network to analyze the spatial distribution characteristics of weather parameters. The input layer receives weather information including precipitation intensity heat maps, temperature gradient distribution, visibility cloud maps, etc., identifies local microclimate areas through a 3×3 convolution kernel, and uses global average pooling to capture global weather patterns such as systematic rainfall belts. The weather propagation paths of adjacent road sections are associated through void convolution, and spatial features are extracted with the help of multi-scale convolution kernels. The output is an environmental spatial distribution feature vector that describes the spatially heterogeneous distribution characteristics of weather parameters.

[0030] Furthermore, the scene attribute feature vector and the environment spatial distribution feature vector are input into the fully connected layer of the multimodal combination neural network for feature fusion. After fusion, the environment scene type and probability distribution are output through the Softmax classifier.

[0031] Specifically, after the two types of feature vectors are input into the fully connected layer, cross-modal feature fusion is achieved through a weight matrix, mapping static scene attributes and dynamic environmental spatial features into a unified feature space. During the fusion process, a nonlinear activation function is used to enhance feature expression and eliminate linear correlations between features, forming a comprehensive feature vector that incorporates multi-dimensional information such as road structure, historical behavior, and weather distribution.

[0032] The fused comprehensive feature vector is processed by the Softmax classifier, and the numerical output is converted into the probability distribution of each environmental scene type. The classifier is based on the preset scene type labels, such as urban congested roads, rural slippery roads, and foggy scenes on highways. The probability value of each scene type is calculated through the normalized exponential function, and the final output includes the distribution results of each scene type and its corresponding probability. In some embodiments, a logistic regression model is established, and specific scene attribute feature vectors, traffic behavior time series feature vectors, and environmental spatial distribution feature vectors are selected and input into the logistic regression model. The historical dominant probability and the scene constraint probability are output. If the historical dominant probability is less than the scene constraint probability, the scene attribute feature vector and the traffic behavior time series feature vector are input into the fully connected layer in the multimodal combination neural network for feature fusion. After fusion, the traffic state type and probability distribution are output through the Softmax classifier.

[0033] Specifically, the scenario attribute feature vector includes static attributes such as road slope, curvature, and number of lanes, as well as historical behavioral features such as the frequency of historical accidents and the proportion of sudden acceleration events. The traffic behavior temporal feature vector encompasses real-time dynamic temporal patterns such as the average vehicle speed over the past five minutes, the standard deviation of acceleration, and the headway distribution. The environmental spatial distribution feature vector includes environmental heterogeneity descriptions such as precipitation intensity heatmaps, visibility gradient distribution, and weather type codes. These specific feature vectors are input into a logistic regression model, trained using a weighted cross-entropy loss function, and analyzed using SHAP values ​​to identify key features, such as the interaction between road friction coefficient and precipitation intensity in slippery road scenarios. The model then weightedly fuses these three features to output a historical dominant probability, reflecting the reliability of predicting traffic conditions based solely on historical behavioral features, and a scenario constraint probability, quantifying the strength of the constraints imposed by the current environmental scenario on traffic conditions. The historical dominant probability reflects the reliability of predicting traffic conditions based solely on historical behavioral features; for example, this probability approaches 1 when a road section has a low historical accident rate and stable speed. The scenario constraint probability quantifies the strength of the constraints imposed by the current environmental scenario on traffic conditions; for example, slippery curves significantly increase this probability. If the historical dominant probability is less than the scenario constraint probability, the scenario attribute feature vector and the traffic behavior time series feature vector are combined through tensor fusion to achieve matrix multiplication interaction to capture nonlinear relationships. The feature weights are then dynamically adjusted through the self-attention mechanism. The fused features are input into the fully connected layer and processed through activation functions such as GeLU and SELU. Finally, the traffic state type and probability distribution are output through the Softmax classifier. In some embodiments, if the historical dominant probability is greater than the scenario constraint probability, the traffic behavior time series feature vector is input into the fully connected layer in the multimodal combination neural network, and the traffic state type and probability distribution are output.

[0034] Specifically, the fully connected layer adopts a dual hidden layer architecture. The first hidden layer contains 256 neurons using the GeLU activation function, and the second hidden layer contains 128 neurons using the SELU activation function. When the input data of the fully connected layer is the fused features, the two hidden layers are used to perform nonlinear transformation on the fused scene attribute feature vector and the traffic behavior time series feature vector. The GeLU activation function can adaptively capture the nonlinear relationship in the features, and the SELU activation function maintains the stability of the feature scale through the self-normalization feature. Finally, through the Softmax classifier, it is mapped into the normalized probability distribution of three types of traffic states: acceleration, deceleration, and uniform speed.

[0035] When the input data of the fully connected layer is only the time series feature vector of traffic behavior, the time series features are nonlinearly transformed through the first hidden layer to extract the basic feature pattern. The features are further processed through the second hidden layer to avoid gradient disappearance. The feature vector transformed by the two hidden layers is input into the Softmax classifier and mapped into the normalized probability distribution of three types of traffic states: acceleration, deceleration, and constant speed.

[0036] In some embodiments, based on the predicted environmental scenarios and traffic conditions, the probability of different travel conditions for vehicles in each sub-area is calculated, specifically including: establishing a random forest model, inputting the environmental scenario type and probability distribution and the traffic condition type and probability distribution into the random forest model, and outputting the travel condition type and probability distribution.

[0037] Specifically, the environmental scene types and probability distributions output by the multimodal combination neural network, such as the probability of urban road scenes at 80% and the probability of high-speed scenes at 20%, as well as the traffic state types and probability distributions, such as the probability of deceleration at 65% and the probability of uniform speed at 30%, are integrated into feature vectors with continuous probability values. For example, the environmental scene types are converted into binary vectors through one-hot encoding, such as [1,0] for urban roads and [0,1] for high speeds, and the probability distribution is directly used as a continuous feature (0.8,0.2); the traffic state types are also one-hot encoded, such as [1,0,0] for deceleration and [0,1,0] for uniform speed, with a probability distribution of (0.65,0.30,0.05). The integrated feature vector contains dimensions such as environmental scene coding, environmental probability, traffic state coding, and traffic probability.

[0038] A random forest model was then built, consisting of multiple decision trees. Each decision tree was constructed by randomly sampling training data and randomly selecting a subset of features. For example, at each split, a subset of features was randomly selected from the environmental scenario type, environmental probability, traffic state type, and traffic probability to reduce model variance. During training, historical travel condition data, such as congested conditions, free-flow conditions, and start-stop conditions, along with their corresponding environmental scenarios and traffic state characteristics, was used as samples. Feature importance was measured using Gini impurity or information gain to determine the splitting rules for the decision tree.

[0039] The environmental scenario type and probability distribution, as well as the traffic state type and probability distribution, of each sub-area are input into the trained random forest model. Each decision tree traverses the feature vector according to the splitting rule obtained through training, generating a prediction of the travel condition type from the root node to the leaf node. For example, for congested conditions, the prediction results of all decision trees are used to determine the final condition type through voting. At the same time, the voting percentage of each condition type is calculated as the probability distribution. For example, if 60 out of 100 decision trees predict congested conditions, 30 predict free-flow conditions, and 10 predict start-stop conditions, the output probability is 60% for congested conditions, 30% for free-flow conditions, and 10% for start-stop conditions. Ultimately, a quantitative result containing the condition type and probability distribution is formed, providing a probabilistic judgment basis for the condition dimension for dynamic driving scenario description.

[0040] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. The scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A vehicle dynamic driving scene determination method based on multi-source information fusion, characterized in that: The method comprises the following steps: Obtaining a target driving path of a target vehicle, and dividing the target driving path into a plurality of continuous sub-areas; Obtain multi-source information parameters for each sub-area and output the estimated environmental scene and traffic status in each sub-area through a multimodal combined neural network; Based on the predicted environmental scenarios and traffic conditions, the probability of vehicles experiencing different travel conditions in each sub-area is calculated; The output includes a dynamic driving scenario description of the environmental scenario, traffic status and operating condition probability.

2. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 1, characterized in that: The obtaining of the target driving path of the target vehicle includes: obtaining a target travel of the target vehicle; and inputting the target travel of the target vehicle into a preset online map path planning model to generate the target driving path of the target vehicle.

3. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 1, characterized in that: The target driving path is divided into a plurality of sub-areas, specifically comprising: presetting a segmentation threshold, and dividing the target driving path into a plurality of sub-areas according to the segmentation threshold.

4. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 1, characterized in that: The multi-source information parameters include high-precision map information, real-time traffic flow information, real-time weather information, and vehicle historical travel information.

5. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 4, characterized in that: The multimodal combined neural network includes a scene attribute feature extraction network, a traffic behavior time series feature extraction network, and an environment space distribution feature extraction network. The scene attribute feature vector of the target sub-area is extracted by the scene attribute feature extraction network, the traffic behavior time series feature vector of the target sub-area is extracted by the traffic behavior time series feature extraction network, and the environment space distribution feature vector of the target sub-area is extracted by the environment space distribution feature extraction network.

6. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 5, characterized in that: The scene attribute feature vector and the environment spatial distribution feature vector are input into the fully connected layer of the multimodal combination neural network for feature fusion. After fusion, the environment scene type and probability distribution are output through the Softmax classifier.

7. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 6, characterized in that: A logistic regression model is established. Specific scene attribute feature vectors, traffic behavior time series feature vectors, and environmental spatial distribution feature vectors are selected as input into the logistic regression model. The historical dominant probability and scene constraint probability are output. If the historical dominant probability is less than the scene constraint probability, the scene attribute feature vector and the traffic behavior time series feature vector are input into the fully connected layer of the multimodal combination neural network for feature fusion. After fusion, the traffic state type and probability distribution are output through the Softmax classifier.

8. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 7, characterized in that: If the historical dominant probability is greater than the scenario constraint probability, the traffic behavior time series feature vector is input into the fully connected layer of the multimodal combination neural network, and the traffic state type and probability distribution are output.

9. The method for determining vehicle dynamic driving scenes based on multi-source information fusion according to claim 8, characterized in that: Based on the predicted environmental scenarios and traffic conditions, the probability of different travel conditions for vehicles in each sub-area is calculated, specifically including: establishing a random forest model, inputting the environmental scenario type and probability distribution as well as the traffic condition type and probability distribution into the random forest model, and outputting the travel condition type and probability distribution.

Images