A pre-warning method and system

Abstract

The present disclosure provides a pre-warning method, system, computer readable storage medium and computer program product, the method comprising: receiving vehicle-to-infrastructure (V2I) information from a roadside unit (RSU), the V2I information comprising characteristic parameters of the RSU and roadside perception data collected by the RSU; inputting the characteristic parameters into a machine learning model to predict a latency of the RSU; compensating a preset pre-warning timing with the latency to obtain a compensated pre-warning timing, the preset pre-warning timing being a preset timing at which a warning should be given when a self-vehicle has a collision risk with other objects; and making a pre-warning judgment based on the compensated pre-warning timing and the roadside perception data.

Description

Technical Field

[0001] This disclosure relates to the field of traffic management, specifically to an early warning method and system. Background Technology

[0002] In scenarios where vehicle collisions are possible, premature warnings from vehicle-to-everything (V2X) applications can negatively impact the driver's experience, while delayed warnings may lead to accidents. Because the latency of Roadside Units (RSUs) varies across different scenarios, and current warning technologies do not account for RSU latency, the actual warning timing of the same V2X application can be later than the required warning timing in different scenarios. Furthermore, the time difference between the actual and required warning timings can differ, affecting the driving experience and potentially causing accidents. For example, on roads with long RSU latency, the V2X application may issue a warning to the driver later than the required warning timing, potentially leading to an accident.

[0003] Therefore, an improved technical solution is urgently needed to solve the above-mentioned problems of the existing technology. Summary of the Invention

[0004] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify the key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as an introduction to the more detailed description that follows.

[0005] This disclosure provides an early warning method, comprising: receiving vehicle-to-infrastructure communication (V2I) information from a roadside unit (RSU), the V2I information including characteristic parameters of the RSU and roadside perception data collected by the RSU; inputting the characteristic parameters into a machine learning model to predict the latency of the RSU; using the latency to compensate for a preset early warning timing to obtain a compensated early warning timing, the preset early warning timing being a preset timing when a vehicle should issue an early warning when there is a risk of collision with other objects; and making an early warning judgment based on the compensated early warning timing and the roadside perception data.

[0006] In some embodiments, the characteristic parameters include at least one of the following: RSU manufacturer, deployment city, identification number (ID), location, information packaging time, number of objects identified, and sensor type.

[0007] In some embodiments, roadside perception data includes: the vehicle ID, location, and speed of the vehicle, and the ID, location, and speed of other objects perceived by the RSU.

[0008] In some embodiments, the compensated warning time is equal to the delay plus the preset warning time.

[0009] In some embodiments, the warning judgment based on the compensated warning timing and roadside perception data further includes: determining the estimated time of collision (TTC) between the vehicle and each of the other surrounding objects based on the roadside perception data, wherein the other surrounding objects include all other objects within the communication range of the vehicle or one or more objects that are closest to the vehicle; and triggering a warning when any estimated TTC is less than or equal to the compensated warning timing.

[0010] In some embodiments, the machine learning model includes a neural network model trained using historical sample data, which includes historical characteristic parameters of multiple sample RSUs and corresponding historical latency data.

[0011] In some embodiments, the machine learning model can be updated based on new sample data.

[0012] In some embodiments, V2I information is broadcast by the RSU at preset time intervals.

[0013] This disclosure also provides a warning system, comprising: an information acquisition module configured to: receive vehicle-to-infrastructure communication (V2I) information from a roadside unit (RSU), the V2I information including characteristic parameters of the RSU and roadside perception data collected by the RSU; a delay prediction module configured to: input the characteristic parameters into a machine learning model to predict the delay of the RSU; a warning module configured to: compensate a preset warning timing using the delay to obtain a compensated warning timing, the preset warning timing being a preset timing when a vehicle should issue a warning when there is a risk of collision with other objects; and make a warning judgment based on the compensated warning timing and the roadside perception data.

[0014] In some embodiments, the characteristic parameters include at least one of the following: RSU manufacturer, deployment city, identification number (ID), location, information packaging time, number of objects identified, and sensor type.

[0015] In some embodiments, roadside perception data includes: the vehicle ID, location, and speed of the vehicle, and the ID, location, and speed of other objects perceived by the RSU.

[0016] In some embodiments, the compensated warning time is equal to the delay plus the preset warning time.

[0017] In some embodiments, the warning module is further configured to make a warning judgment based on the compensated warning timing and roadside perception data by: determining the estimated time of collision (TTC) between the vehicle and each of the other objects in the surrounding area based on the roadside perception data, the other objects in the surrounding area including all other objects within the communication range of the vehicle or one or more objects closest to the vehicle; and triggering a warning when any estimated TTC is less than or equal to the compensated warning timing.

[0018] In some embodiments, the machine learning model includes a neural network model trained using historical sample data, which includes historical characteristic parameters of multiple sample RSUs and corresponding historical latency data.

[0019] In some embodiments, the machine learning model can be updated based on new sample data.

[0020] In some embodiments, V2I information is broadcast by the RSU at preset time intervals.

[0021] This disclosure also provides a computer-readable storage medium storing a computer program for early warning, which can be executed by a processor to perform the aforementioned early warning method.

[0022] This disclosure also provides a computer program product, including instructions that can be executed by a processor to perform the aforementioned warning method. Attached Figure Description

[0023] The features, nature, and advantages of this disclosure will become more apparent when understood in conjunction with the accompanying drawings. It should be noted that the described drawings are schematic and not limiting. In the drawings, some parts are enlarged and are not drawn to scale for illustrative purposes.

[0024] Figure 1 The diagram illustrates the warnings for both uncompensated and compensated RSU delay scenarios.

[0025] Figure 2 The early warning method of this disclosure is shown.

[0026] Figure 3 An equivalent schematic diagram of the machine learning model used in this disclosure is shown.

[0027] Figure 4 An exemplary training and prediction process for the machine learning model of this disclosure is shown.

[0028] Figure 5 An example of early warning judgment based on compensated early warning timing and roadside perception data is shown in this disclosure.

[0029] Figure 6A schematic hierarchical structure of the OBU (On-Board Unit) of this disclosure is shown.

[0030] Figure 7 An exemplary data stream of the warnings provided in this disclosure is shown.

[0031] Figure 8 A block diagram of the early warning system of this disclosure is shown. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the described exemplary embodiments. However, it will be apparent to those skilled in the art that the described embodiments can be practiced without some or all of these specific details. In other exemplary embodiments, well-known structures have not been described in detail to avoid unnecessarily obscuring the concepts of this disclosure. It should be understood that the specific embodiments described herein are merely illustrative of this disclosure and are not intended to limit it. Furthermore, the various aspects described in the embodiments can be combined arbitrarily without conflict.

[0033] As mentioned above, existing warning schemes do not consider the delay of the RSU (Roadside Unit), which can affect the driving experience and may even lead to safety accidents. In view of this, this disclosure considers the RSU delay when determining the warning timing, uses a machine learning model based on the characteristic parameters of the RSU to predict the RSU delay, and uses the predicted delay to compensate for the preset warning timing so that the actual warning timing is close to the appropriate warning timing.

[0034] Warning timing can be expressed in various forms, such as an absolute time point, a time interval from the predicted collision time, or a time interval from the current time. For example, when using an absolute time point, the warning timing can be represented by a specific time point (such as X hours, Y minutes, Z seconds), indicating that a warning will be issued at that specific time point. When using a time interval from the predicted collision time, the warning timing can be a certain time interval (such as 5 seconds), indicating that a warning will be issued 5 seconds before the predicted collision. And when using a time interval from the current time, the warning timing can be a certain time interval (such as 5 seconds), indicating that a warning will be issued 5 seconds after the current time.

[0035] For ease of explanation, this disclosure uses a time period from the expected collision time to represent the warning timing. Specifically, the time when it is predicted that the vehicle will collide with another object without intervention can be used to represent the warning timing, and the time period between the issuance of the warning and the expected collision time can be used. It should be noted that the form of representing the warning timing used in this disclosure is merely exemplary and not limiting. In specific implementations, those skilled in the art can use other forms (e.g., absolute time points, time periods from the current time) to represent the warning timing according to the actual situation.

[0036] Figure 1 The diagram illustrates the warnings for both uncompensated and compensated RSU delay scenarios.

[0037] In this disclosure, RSU latency refers to the time interval between the time the RSU collects data and the time the on-board unit (OBU) receives the data from the RSU. Specifically, RSU latency includes not only the time interval for data to be sent from the RSU to the OBU, but also the time interval for the RSU to process the data after collection (e.g., filtering, packaging, etc.). For example, if the RSU collects data at time A, processes the collected data, and sends it to the OBU at time B, and the OBU receives the data from the RSU at time C, then the RSU latency is the time interval (C-A).

[0038] Figure 1 The left side of the diagram illustrates a warning scenario without RSU delay compensation. During vehicle operation, the vehicle (HV) continuously detects collision risks with surrounding objects (e.g., other vehicles, RVs). When the HV detects a collision risk with another vehicle (e.g., the distance between them is too close), it should issue a warning before the actual collision occurs, allowing timely intervention to avoid it. In this disclosure, the term "warning time" is used to indicate the ideal timing when the HV detects a collision risk with another vehicle without RSU delay. If the actual warning time is no later than the warning time, the HV has sufficient time to take action to avoid a collision. Conversely, if the actual warning time is later than the warning time, the HV may not have sufficient time to take action, thus posing a collision risk.

[0039] In the context of this disclosure, "the actual warning timing is later than the required warning timing" means that the starting point of the time period corresponding to the actual warning timing is later than the starting point of the time period corresponding to the required warning timing. This means that the actual warning was issued later than the time when a warning should ideally be issued.

[0040] like Figure 1As shown on the left, without compensation for RSU delay, the actual warning timing will be later than the expected warning timing due to the RSU delay. For example, suppose the expected warning timing is when the estimated time to collision (TTC) between the vehicle and another vehicle is equal to 5 seconds. TTC refers to the time required for the vehicle to collide with another object while maintaining the current speed and direction. Specifically, TTC can be expressed as follows:

[0041]

[0042] TTC (Traffic Traction Control) can be used to assess a vehicle's collision risk. When the TTC falls below a certain threshold, it indicates that the vehicle is too close to other objects, posing a collision risk. Therefore, the vehicle can determine whether a warning is needed by comparing the calculated TTC value with the threshold.

[0043] Continuing the previous example, suppose the warning should be issued when the Time-to-Traffic (TTC) between the vehicle and other vehicles equals 5 seconds, and assume the Responsive Safety Unit (RSU) delay is 1 second. In this case, due to the RSU delay, the information the vehicle receives from the RSU is delayed, resulting in a deviation in calculating the distance and relative speed between the vehicle and other objects, leading to an inaccurate TTC. For example, when the vehicle calculates a TTC of 5 seconds, the actual TTC may be 4 seconds (calculated TTC minus delay). Therefore, when the vehicle makes a warning decision based on the calculated TTC, it will issue a warning when the calculated TTC equals 5 seconds (actual TTC equals 4 seconds), later than the expected warning time (when the TTC equals 5 seconds), creating a dangerous situation.

[0044] To compensate for the impact of RSU latency, the technical solution disclosed herein considers RSU latency when making early warning judgments. Accordingly, Figure 1 The right side shows a diagram illustrating the early warning system under the condition of RSU delay compensation.

[0045] Similarly, suppose the warning should be issued when the TTC between the vehicle and other vehicles equals 5 seconds, and assume the RSU delay is 1 second. In this case, the warning timing can be compensated to obtain a compensated warning timing. Specifically, the compensated warning timing can be equal to the required warning timing plus the RSU delay (i.e., 5 seconds + 1 second = 6 seconds). Therefore, the vehicle can make a warning judgment based on the compensated warning timing. The vehicle can issue a warning when the calculated TTC equals 6 seconds. Due to the RSU delay, when the vehicle's calculated TTC equals 6 seconds, the actual TTC equals 5 seconds. It is evident that when the vehicle makes a warning judgment based on the compensated warning timing, the actual warning timing is later than the compensated warning timing but closer to the required warning timing due to the RSU delay, resulting in a safe situation. Figure 1 As shown on the right.

[0046] pass Figure 1 It is evident that RSU latency has a significant impact on early warning timing. To more clearly demonstrate how this disclosure accurately predicts latency and achieves adaptive early warning through machine learning models, the following combines... Figure 2 The overall process of the early warning method disclosed herein is explained in detail.

[0047] Specifically, Figure 2 A warning method 200 of this disclosure is shown. The method 200 can be executed by a vehicle (e.g., the vehicle's OBU).

[0048] like Figure 2 As shown, method 200 begins at step 205. In step 205, vehicle-to-infrastructure communication (V2I) information is received from a roadside unit (RSU), which includes characteristic parameters of the RSU and roadside sensing data collected by the RSU.

[0049] In some embodiments, whether the RSU is within the vehicle's communication range can be determined based on distance or actual signal strength. For example, when the distance between the vehicle and the RSU does not exceed a preset distance threshold, the vehicle and the RSU are considered to be within their respective communication ranges. Similarly, when the distance between the vehicle and another vehicle does not exceed a preset distance threshold, the vehicle and the other vehicle are considered to be within their respective communication ranges. In other embodiments, whether the RSU is within the vehicle's communication range can be determined based on actual signal strength. For example, when the signal transmitted between the vehicle and the RSU meets a preset strength threshold, the vehicle and the RSU are considered to be within their respective communication ranges. Similarly, when the signal transmitted between the vehicle and another vehicle meets a preset strength threshold, the vehicle and the other vehicle are considered to be within their respective communication ranges.

[0050] In some embodiments, the characteristic parameters of the RSU include, but are not limited to, at least one of the following: RSU manufacturer, deployment city, identification number (ID), location, information packaging time, number of objects identified, and sensor type.

[0051] Among the RSU's characteristics mentioned above, "Manufacturer" indicates the manufacturer responsible for producing the RSU. "Deployment City" indicates the city within its region where the RSU is deployed. "RSU ID" uniquely identifies the RSU. "Location" indicates the geographical location of the RSU (e.g., GPS coordinates). "Information Packaging Time" indicates the time required for the RSU to process and package information; this parameter is included with the RSU at the factory. "Object Recognition Count" indicates the maximum number of objects / surfaces the RSU can simultaneously recognize. "Sensor Type" indicates the type of sensors used by the RSU, such as cameras, LiDAR, etc.

[0052] In some embodiments, the roadside perception data collected by the RSU includes, but is not limited to: the vehicle ID, location, and speed of the vehicle, as well as the ID, location, and speed of other objects perceived by the RSU.

[0053] RSUs can sense relevant information about objects (vehicles, pedestrians, and / or obstacles) within their communication range, including the object's ID, location, and speed. Therefore, roadside perception data can include the vehicle's ID, location, and speed, as well as the IDs, locations, and speeds of other objects (vehicles, pedestrians, and / or obstacles other than the vehicle itself).

[0054] By collecting and broadcasting roadside sensing data from the RSU (Roadside Sense Unit), the autonomous vehicle can receive this data and determine its relative position and speed to other objects, distinguishing them by their IDs. This allows the autonomous vehicle to accurately understand its surrounding traffic conditions and precisely assess collision risks.

[0055] In some embodiments, V2I information can be broadcast by the RSU at preset time intervals. By broadcasting, all vehicles within the RSU's communication range can receive the V2I information. Furthermore, the broadcast time interval can be appropriately set to ensure that vehicles can obtain the latest V2I information in a timely manner. Based on the latest V2I information broadcast by the RSU, the vehicle can issue warnings, enabling it to promptly capture the latest changes in traffic conditions (such as vehicle acceleration, deceleration, lane changes, etc.), thus improving the accuracy and real-time performance of the warnings.

[0056] After obtaining the characteristic parameters of the RSU and roadside perception data through V2I information, method 200 proceeds to step 210. In step 210, the characteristic parameters are input into a machine learning model to predict the RSU latency.

[0057] Different characteristic parameters can have varying impacts on RSU latency. For example, different manufacturers may use different technical standards and hardware configurations when producing RSUs, resulting in differences in latency performance. Traffic conditions, environmental factors, infrastructure layout, and RSU deployment requirements may differ between cities, affecting RSU latency. Different ID numbers for RSUs from the same manufacturer represent different models, and different models may have different latency performance. The location of the RSU affects its communication distance and signal strength with vehicles, thus impacting latency. The longer the RSU's information packaging time, the greater its latency. The more objects the RSU identifies, the more information it needs to process, potentially increasing latency. Furthermore, differences in data acquisition and processing speeds between different types of sensors can also affect RSU latency.

[0058] Machine learning models can be implemented using various models, such as neural network models, decision tree models, and so on. By way of example and not limitation, this disclosure uses a neural network model to implement a machine learning model.

[0059] To better understand this model, Figure 3 An equivalent schematic diagram of the machine learning model of this disclosure is shown. For example... Figure 3 As shown, the machine learning model is implemented by a neural network model that includes an input layer, a hidden layer, and an output layer.

[0060] The input layer is used to receive the characteristic parameters x from the RSU. In some examples, the input layer can also perform certain preprocessing on the received characteristic parameters, such as data cleaning, normalization, etc.

[0061] Hidden layers are the core component of a neural network model, used to process input data to extract features and make predictions based on them. The number of hidden layers and the specific structure of each hidden layer can be set and adjusted according to actual needs to achieve the best prediction results.

[0062] The output layer is used to output the RSU delay in seconds T predicted by the neural network model based on the input characteristic parameters.

[0063] like Figure 3 As shown, the neural network model can be equivalent to the function f(), where the input to the function is the characteristic parameter x of RSU, and the output is the delay in seconds T. The output T of the function is related to the input characteristic parameter x and the intermediate weight value a of each characteristic parameter: T = f(a, x).

[0064] By training the model, the value of the intermediate weight 'a' can be determined, thus obtaining the trained machine learning model. Specifically, the known characteristic parameter 'x' and the corresponding latency 'T' obtained from experiments can be input into the machine learning model for training, and the value of the intermediate weight 'a' can be obtained through reverse calculation.

[0065] After training the machine learning model, in subsequent use, since the intermediate weight value 'a' has been obtained, when a new feature parameter 'x' is acquired, the trained model can predict the corresponding RSU latency.

[0066] After understanding the equivalent model of a machine learning model, the following combines... Figure 4 This further demonstrates an exemplary training and prediction process for machine learning models.

[0067] Figure 4 The upper part shows an exemplary training process for a machine learning model, and the lower part shows an exemplary prediction process for a machine learning model.

[0068] Before model training begins, a large amount of training data needs to be prepared. In various embodiments of this disclosure, historical sample data can be used to train the machine learning model. Historical sample data includes historical characteristic parameters of multiple sample RSUs and corresponding historical latency data, such as... Figure 4 As shown in the image.

[0069] Specifically, multiple RSU samples with different values ​​of characteristic parameters (e.g., different manufacturers, different cities, different sensor types, etc.) can be selected for experiments to obtain the latency T corresponding to each value of the characteristic parameter. These characteristic parameters and corresponding latency data are then used as historical sample data. Furthermore, the data obtained from the experiments can be preprocessed, such as through data filtering and cleaning, to improve the quality of the training data. Abundant historical data provides sufficient and diverse learning samples for model training, enabling the model to accurately predict RSU latency.

[0070] As mentioned earlier, different characteristic parameters may have varying impacts on RSU latency. Among these parameters, some may have a significant impact on RSU latency, while others may have a relatively smaller impact. However, it is currently difficult to quantify the specific degree of impact of each characteristic parameter on RSU latency. Therefore, this disclosure inputs all types of characteristic parameters into the machine learning model for training, rather than selecting only a subset. By comprehensively inputting these characteristic parameters that may affect RSU latency into the machine learning model, the relationship between various factors and latency can be fully explored, greatly improving the accuracy of RSU latency prediction and providing a more reliable basis for subsequent early warning timing compensation.

[0071] In addition, the model needs to be initialized. For example, the intermediate weight values ​​'a' can be initialized. In some examples, the initial weight values ​​can be randomly generated. In other examples, specific initialization methods, such as Xavier initialization or He initialization, can be used to improve the model's convergence speed.

[0072] After preparing the training data and initializing the model, training can begin. During training, the model's input layer receives historical characteristic parameters and predicts latency data. The loss function is calculated by comparing the predicted latency data with the known historical latency data. Intermediate weight values ​​'a' are iteratively updated using an optimization algorithm (e.g., gradient descent or other optimization algorithms) to reduce the loss function. During this process, the intermediate weight values ​​'a' will change. If a certain characteristic parameter has a significant impact on RSU latency, the change in the intermediate weight value 'a' will be highly correlated with that characteristic parameter; conversely, if a certain characteristic parameter has a small impact on RSU latency, the change in the intermediate weight value 'a' will be less correlated with that characteristic parameter. When the loss function meets a preset convergence condition after multiple iterations (e.g., the value of the loss function decreases to a certain preset threshold), the training process is complete, resulting in the trained machine learning model. It should be noted that the model training process shown above is merely an example. In practice, those skilled in the art can employ different training processes to train machine learning models. The specific training details of machine learning models (such as model initialization methods, loss function selection, loss function convergence condition settings, optimization algorithm selection, etc.) are well known in the field of machine learning and will not be elaborated here.

[0073] The trained machine learning model can be integrated into the vehicle (e.g., an onboard unit) to predict the latency of the Remote Substation (RSU). Specifically, each time the vehicle passes an RSU (whether it's the same RSU or a different one), it can receive characteristic parameters broadcast by that RSU. For example... Figure 4 As shown in the lower part, the vehicle can input the new characteristic parameters received into the trained machine learning model, and the machine learning model then predicts the RSU latency based on the RSU characteristic parameters and outputs the prediction result.

[0074] After the trained machine learning model is deployed to various vehicles, these vehicles generate new data during model usage. In some embodiments, this data can be used as new sample data to update the machine learning model. For example, each vehicle can upload the new sample data to a server, which uses this data to iterate and update the machine learning model, resulting in an updated model. The server can then distribute the updated model to each vehicle. In this way, the machine learning model can be continuously optimized during use, thereby improving the predictive performance of RSU latency and ultimately enhancing the accuracy and reliability of warnings.

[0075] return Figure 2After predicting the RSU delay, method 200 proceeds to step 215. In step 215, the delay is used to compensate for the preset warning timing to obtain a compensated warning timing, which is a preset timing when a warning should be issued when there is a risk of collision between the vehicle and other objects.

[0076] In various embodiments of this disclosure, the preset warning timing is a pre-set fixed time period (e.g., 5 seconds). For example, when the vehicle detects a collision risk with another object, the vehicle issues a warning 5 seconds before the expected collision. In the context of this disclosure, "preset warning timing" is equivalent to the above-mentioned reference. Figure 1 The term "warning time" is described. In an ideal situation without RSU delay, the actual warning time is close to the preset warning time, and the situation is safe. However, in the presence of RSU delay, the actual warning time is later than the preset warning time, and the situation is dangerous.

[0077] In the various embodiments of this disclosure, the compensated warning timing is equal to the delay plus the preset warning timing. For example, assuming the preset warning timing is 5 seconds and the predicted RSU delay is 1 second, the compensated warning timing is 6 seconds. The compensated warning timing calculated in this way fully considers the RSU delay, ensuring that subsequent warning judgments based on the compensated warning timing compensate for the deviation caused by the RSU delay, thus ensuring that the actual warning timing is close to the appropriate warning timing.

[0078] In step 220, a warning judgment is made based on the compensated warning timing and roadside perception data.

[0079] After obtaining the compensated warning timing and roadside perception data, the estimated time of collision (TTC) between the vehicle and each of the surrounding objects can be determined based on the roadside perception data. These surrounding objects include all other objects within the vehicle's communication range or one or more objects closest to the vehicle. A warning can then be triggered when any estimated TTC is detected to be less than or equal to the compensated warning timing.

[0080] To better understand the process of early warning judgment, the following combines... Figure 5 To illustrate an example of early warning judgment, for ease of explanation, Figure 5 It is assumed that there is only one other object (another vehicle RV) within the communication range of the vehicle (HV), the RSU delay is 1 second, the preset warning time is 5 seconds, and the compensated warning time is 6 seconds.

[0081] During operation, the vehicle can calculate its own Time Toll Collection (TTC) and that of other vehicles based on roadside perception data. When the calculated TTC is less than or equal to the compensated warning time, the vehicle can trigger a warning. Due to the RSU delay, the TTC calculated by the vehicle (referred to as "estimated TTC" in this disclosure) is not the actual TTC. Specifically, actual TTC = estimated TTC - delay.

[0082] like Figure 5 As shown, at time t1, the estimated TTC calculated by the vehicle is 8 seconds. Since the estimated TTC (8 seconds) is greater than the compensated warning timing (6 seconds), the vehicle does not trigger a warning at time t1.

[0083] At time t2, following time t1, the vehicle calculates an estimated Time To Collision (TTC) of 6 seconds. Since the estimated TTC (6 seconds) equals the compensated warning timing (6 seconds), the vehicle triggers a warning at time t2. At time t2, the actual TTC = estimated TTC (6 seconds) - delay (1 second) = 5 seconds. That is, the vehicle actually issues a warning 5 seconds before the expected collision, close to the preset warning timing, indicating a safe situation.

[0084] For the warning application in the vehicle's OBU, the input to the warning application is the compensated warning timing, not the preset warning timing. When making a warning judgment, the warning application determines whether to trigger the warning based on a comparison between the estimated TTC calculated by the vehicle and the compensated warning timing. The estimated TTC calculated by the vehicle is affected by the RSU delay, while the compensated warning timing, which utilizes RSU compensation, offsets the impact of the RSU delay, thus enabling the warning application to trigger the warning at the accurate time.

[0085] By comparing the estimated TTC with the compensated warning timing to trigger the warning, the impact of RSU delay is effectively offset, making the actual warning timing close to the required warning timing, improving the accuracy of the warning, effectively reducing the risk of collision, and enhancing driving safety.

[0086] In the various embodiments of this disclosure, warnings can be issued in multiple ways, such as sound, text, and video. Sound warnings can quickly attract the driver's attention through sound. For example, a specific alarm sound can be emitted through the vehicle's speakers. Additionally, specific risk situations can be communicated via voice prompts, such as "The vehicle ahead is too close, please slow down." Text warnings can display specific hazard information on the vehicle's display screen, such as the location of the potential collision object, the predicted collision time, and suggested avoidance measures. For example, the display screen can show "The vehicle ahead is too close, the estimated collision time is 5 seconds, please slow down and maintain a safe distance." Text warnings can provide the driver with more detailed information, helping the driver better understand the collision risk. Video warnings can display simulated animations of collision risks. By displaying the relative position and trajectory of the vehicle and surrounding objects on the vehicle's display screen, the driver can intuitively see the potential collision risk. For example, video warnings can use simulated animations to display the trajectory of the vehicle and other vehicles and the expected collision point, thereby helping the driver to make judgments and reactions in advance.

[0087] In practical applications, the appropriate warning method can be selected based on the actual situation. For example, for minor collision risks, the vehicle can issue a low-volume audible warning; while for serious collision risks, the vehicle can issue a high-volume audible warning, and / or simultaneously issue at least two of the following warning methods: audible, text, and video warnings. Furthermore, drivers can personalize their warning settings according to their driving habits and preferences. For example, they can choose to receive only a single type of warning (e.g., an audible warning) or a combination of multiple warning methods.

[0088] In some embodiments, after a warning is triggered, the driver of the vehicle can take appropriate obstacle avoidance measures based on the warning. In other embodiments, the vehicle can also automatically take certain auxiliary measures (e.g., automatically change driving speed, automatically change driving direction, etc.) to assist the driver in avoiding obstacles.

[0089] The above describes the scenario where there is only one other object within the vehicle's communication range. When there are multiple other objects within the vehicle's communication range (e.g., five other objects), the vehicle can select one or more objects that are closest to it (e.g., the three closest other objects) and determine the Time To Call (TTC) between the vehicle and each selected other object based on roadside perception data (i.e., three TTCs). A warning can be triggered when any one of these three TTCs is less than or equal to the compensated warning timing. Alternatively, the vehicle can select all other objects within its communication range (e.g., five other objects) and determine the TTC between the vehicle and each other object based on roadside perception data (i.e., five TTCs). A warning can be triggered when any one of these five TTCs is less than or equal to the compensated warning timing.

[0090] In practice, the vehicle can choose to monitor the TTC of all other objects within its communication range or a subset of those objects, depending on its own computing resources. For example, if the vehicle has sufficient computing resources, it can choose to monitor the TTC of all other objects within its communication range. However, if computing resources are limited, it can choose to monitor the TTC of the subset of other objects within its communication range that are closest to the vehicle.

[0091] As can be seen, when there are multiple other objects around the vehicle, the vehicle can monitor the TTC of all other surrounding objects, as well as the TTC of several other objects closest to the vehicle. This approach improves the flexibility of the early warning scheme.

[0092] Under normal circumstances, the vehicle operates only within the communication range of a single Roadside Utility Unit (RSU). In such cases, the vehicle obtains characteristic parameters and roadside perception data from the same RSU, predicts the latency of that RSU based on this data, and compensates for the latency of that RSU to make early warning judgments.

[0093] In extreme cases, the vehicle may simultaneously be within the communication range of multiple Roadside Units (RSUs) (e.g., two RSUs) (e.g., the vehicle is in the overlapping area of ​​the communication ranges of two RSUs). In such cases, the vehicle can take the characteristic parameters of one RSU and roadside perception data to predict the latency of the same RSU, and compensate for the latency of the same RSU to make a warning judgment. That is, the same RSU is used in each step of the warning method 200. In this way, the confusion and conflict that may occur in the case of multiple RSUs can be avoided, ensuring that the warning scheme of this disclosure can operate stably in complex scenarios.

[0094] Method 200 can be continuously executed by the vehicle while it is in motion. In this way, the latest V2I information can be continuously obtained, and the RSU latency can be predicted and compensated based on the latest V2I information to provide early warning at the right time.

[0095] As can be seen from Method 200, using a machine learning model to predict the RSU delay based on characteristic parameters, compensating for the preset warning timing with the predicted delay, and making a warning judgment based on the compensated warning timing can make the actual warning timing close to the appropriate warning timing. In this way, drivers can be warned at the appropriate time, effectively avoiding safety accidents caused by late warnings and improving driving safety. At the same time, it also avoids affecting the driving experience due to premature warnings, achieving precise control of the warning timing and balancing driving safety and driving experience.

[0096] The following will combine Figures 6 to 8 The technical solution disclosed herein will be further explained from the perspectives of layered structure, data flow, and system architecture.

[0097] Figure 6 A schematic hierarchical structure of the OBU of this disclosure is shown.

[0098] As shown in the figure, the layered structure of the OBU includes the system layer, the facility layer, and the application layer.

[0099] The system layer is located at the bottom of the layered structure, providing the basic support required for OBU operation, such as computing resources, storage, access control, network connectivity, fault recovery, and so on.

[0100] The facility layer is located above the system layer and is mainly used to handle matters related to V2I communication, such as receiving V2I information from the RSU and analyzing V2I information.

[0101] The application layer sits above the infrastructure layer and is mainly used to handle matters related to V2X applications, such as using information provided by the infrastructure layer for early warning and user interaction.

[0102] like Figure 6 As shown, the V2I information received from the RSU resides in the infrastructure layer. The machine learning model is located between the infrastructure layer and the application layer. The machine learning model can predict the RSU latency based on the V2I information and provide it to V2X applications in the application layer (e.g., warning applications). The V2X application makes warning decisions based on the latency output by the machine learning model and the roadside perception data contained in the V2I information.

[0103] It should be noted that Figure 6 The OBU layered structure shown is only an exemplary architecture. In practice, OBUs can be designed with other layered structures according to different needs and scenarios.

[0104] The layered structure of the OBU provides underlying support for the early warning function; however, in actual operation, dynamic data interaction also plays a crucial role. Therefore, Figure 7 An exemplary data stream 700 of the warning provided in this disclosure is shown, which reveals the complete data link for triggering the warning.

[0105] like Figure 7 As shown, the V2I information provided by the RSU is the starting point of the entire data stream. The V2I information includes the characteristic parameters of the RSU and the roadside sensing data collected by the RSU.

[0106] The characteristic parameters are then fed into a machine learning model, which uses the characteristic parameters to predict the RSU latency and outputs the predicted latency data.

[0107] Latency data and roadside perception data can be input into vehicular V2X applications (e.g., warning applications). In some embodiments, to further improve the reliability and stability of warnings, the vehicle itself can also input its own driving data into the vehicular V2X application.

[0108] V2X applications use the received data to calculate warning data. Warning data may include, for example, compensated warning timing and the vehicle's time-to-market (TTC) with other objects. V2X applications can use the aforementioned method 200 to compensate for preset warning timing to obtain compensated warning timing. Simultaneously, V2X applications can also calculate the vehicle's TTC with other objects based on roadside perception data.

[0109] Finally, the V2X application can provide warning data to the HMI (Human-Machine Interface). The HMI then makes a warning decision based on the warning data. For example, the HMI can trigger a warning when any TTC is less than or equal to the compensated warning time. In other embodiments, the V2X application can also directly make the warning decision. When the V2X application triggers a warning, it can display warning-related information on the HMI, such as the location of the collision object and the predicted collision time.

[0110] Figure 8 A block diagram of the warning system 800 of this disclosure is shown. In various embodiments of this disclosure, the warning system 800 may be installed in a vehicle (e.g., in the vehicle's OBU).

[0111] See Figure 8 The system 800 may include an information acquisition module 805, a delay prediction module 810, and an early warning module 815. As shown in the figure, the information acquisition module 805, the delay prediction module 810, and the early warning module 815 may be directly or indirectly connected to or communicate with each other on one or more buses 820.

[0112] In various embodiments of this disclosure, the information acquisition module 805 is configured to receive vehicle-to-infrastructure communication (V2I) information from a roadside unit (RSU), the V2I information including characteristic parameters of the RSU and roadside perception data collected by the RSU.

[0113] The delay prediction module 810 receives characteristic parameters from the information acquisition module 805 and is configured to input the characteristic parameters into a machine learning model to predict the delay of RSU.

[0114] The warning module 815 receives roadside perception data from the information acquisition module 805 and time delay from the time delay prediction module 810 and is configured to: compensate for the preset warning timing using the time delay to obtain a compensated warning timing, wherein the preset warning timing is a preset timing when the vehicle has a risk of collision with other objects and should issue a warning; and make a warning judgment based on the compensated warning timing and the roadside perception data.

[0115] In some embodiments, the warning module 815 is further configured to: determine the estimated time of collision (TTC) between the vehicle and each of the other objects in the surrounding area based on roadside perception data, the other objects in the surrounding area including all other objects within the communication range of the vehicle or one or more objects closest to the vehicle; and trigger a warning when any estimated TTC is less than or equal to the compensated warning timing.

[0116] In various embodiments of this disclosure, the information acquisition module 805, the delay prediction module 810, and the early warning module 815 can be implemented in software (e.g., by running corresponding program code to complete their respective functions), in hardware (e.g., by using hardware devices such as application-specific integrated circuits (ASICs) or field-programmable gate arrays (FPGAs) to implement the functions of the modules), or in a combination of software and hardware.

[0117] Figure 8 Specific modules of the warning system 800 are shown, but it should be understood that these modules are exemplary and not limiting. In different implementations, one or more of these modules may be combined, split, removed, or additional modules may be added. For example, in some implementations, system 800 may also add an output module for outputting alert information related to the warning.

[0118] The detailed description above, in conjunction with the accompanying drawings, describes examples but does not represent all examples that can be implemented or fall within the scope of the claims. The terms "example" and "exemplary" are used in this specification to mean "serving as an example, instance, or illustration" and do not imply "superiority or superiority over other examples."

[0119] Throughout this specification, the terms "an embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of this disclosure. Therefore, the use of these phrases may refer to more than one embodiment. Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0120] The preceding description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will readily be understood by those skilled in the art, and the universal principles defined herein can be applied to other aspects. Therefore, the claims are not intended to be limited to the aspects shown herein, but are to be granted the full scope consistent with the language of the claims, wherein references to the singular form of an element, unless specifically stated otherwise, are not intended to mean “one and only one,” but rather “one or more.” Unless specifically stated otherwise, the term “some” refers to one or more. All structural and functional equivalents of the elements of the various aspects described throughout this disclosure, whether now or hereafter known to a person skilled in the art, are expressly incorporated herein by reference and are intended to be covered by the claims.

[0121] It should also be noted that these embodiments may be described as processes depicted as flowcharts, flow diagrams, structure diagrams, or block diagrams. Although a flowchart may describe the operations as a sequential process, many of these operations can be executed in parallel or concurrently. Furthermore, the order of these operations can be rearranged.

[0122] While various embodiments have been described and illustrated, it should be understood that the embodiments are not limited to the precise configurations and components described above. Various modifications, substitutions, and improvements that will be apparent to those skilled in the art can be made to the arrangement, operation, and details of the apparatus disclosed herein without departing from the scope of the claims.

Claims

1. An early warning method, comprising: Receive vehicle-to-infrastructure (V2I) communication information from a roadside unit (RSU), the V2I information including characteristic parameters of the RSU and roadside sensing data collected by the RSU; The characteristic parameters are input into a machine learning model to predict the RSU latency; The time delay is used to compensate for the preset warning timing to obtain a compensated warning timing. The preset warning timing is a timing when the vehicle should issue a warning when there is a risk of collision with other objects. as well as The warning judgment is made based on the compensated warning timing and the roadside perception data.

2. The method according to claim 1, characterized in that, The characteristic parameters include at least one of the following: the manufacturer of the RSU, the city of deployment, the identification number (ID), the location, the information packaging time, the number of objects identified, and the sensor type.

3. The method according to claim 1, characterized in that, The roadside sensing data includes: the vehicle ID, location, and speed of the vehicle, as well as the ID, location, and speed of other objects sensed by the RSU.

4. The method according to claim 1, characterized in that, The compensated early warning timing is equal to the delay plus the preset early warning timing.

5. The method according to claim 4, characterized in that, The early warning judgment based on the compensated early warning timing and the roadside sensing data further includes: Based on the roadside perception data, the estimated time of collision (TTC) between the vehicle and each of the surrounding objects is determined. These surrounding objects include all other objects within the vehicle's communication range or one or more objects closest to the vehicle. An alert is triggered when any estimated TTC is less than or equal to the compensated alert timing.

6. The method according to claim 1, characterized in that, The machine learning model includes a neural network model trained using historical sample data, which includes historical characteristic parameters of multiple sample RSUs and corresponding historical latency data.

7. The method according to claim 6, characterized in that, The machine learning model can be updated based on new sample data.

8. The method according to claim 1, characterized in that, The V2I information is broadcast by the RSU at preset time intervals.

9. An early warning system, comprising: The information acquisition module is configured to receive vehicle-to-infrastructure communication (V2I) information from a roadside unit (RSU), the V2I information including characteristic parameters of the RSU and roadside sensing data collected by the RSU; A latency prediction module is configured to input the characteristic parameters into a machine learning model to predict the latency of the RSU; The early warning module is configured as follows: The time delay is used to compensate for the preset warning timing to obtain a compensated warning timing. The preset warning timing is a timing when the vehicle should issue a warning when there is a risk of collision with other objects. as well as The warning judgment is made based on the compensated warning timing and the roadside perception data.

10. The system according to claim 9, characterized in that, The characteristic parameters include at least one of the following: the manufacturer of the RSU, the city of deployment, the identification number (ID), the location, the information packaging time, the number of objects identified, and the sensor type.

11. The system according to claim 9, characterized in that, The roadside sensing data includes: the vehicle ID, location, and speed of the vehicle, as well as the ID, location, and speed of other objects sensed by the RSU.

12. The system according to claim 9, characterized in that, The compensated early warning timing is equal to the delay plus the preset early warning timing.

13. The system according to claim 12, characterized in that, The early warning module is further configured to make an early warning judgment based on the compensated early warning timing and the roadside sensing data through the following operations: Based on the roadside perception data, the estimated time of collision (TTC) between the vehicle and each of the other surrounding objects is determined. The other surrounding objects include all other objects within the communication range of the vehicle or one or more objects that are closest to the vehicle. as well as An alert is triggered when any estimated TTC is less than or equal to the compensated alert timing.

14. The system according to claim 9, characterized in that, The machine learning model includes a neural network model trained using historical sample data, which includes historical characteristic parameters of multiple sample RSUs and corresponding historical latency data.

15. The system according to claim 14, characterized in that, The machine learning model can be updated based on new sample data.

16. The system according to claim 9, characterized in that, The V2I information is broadcast by the RSU at preset time intervals.

17. A computer-readable storage medium storing a warning computer program, said computer program being executable by a processor to perform the method according to any one of claims 1 to 8.

18. A computer program product comprising instructions executable by a processor to perform the method according to any one of claims 1 to 8.

Images