Prediction model training method and device, and parking type prediction method and device

Abstract

Embodiments of the present application provide a prediction model training method and device, the method comprising: obtaining a first speed curve and a first relative distance curve corresponding to a driving process of a vehicle. According to the first speed curve, at least one parking event in the driving process is determined, and a first parking position corresponding to the parking event is obtained. For any parking event, according to the first speed curve and a parking duration threshold, a real parking type of the parking event is labeled. According to a prediction model, the first speed curve, the first relative distance curve and the first parking position are processed to obtain a predicted parking type output by the prediction model. According to the real parking type and the predicted parking type, model parameters of the prediction model are updated. Through training, a prediction model capable of outputting an accurate predicted parking type is obtained, so that the internal combustion engine is turned off based on the start-stop system only when the predicted parking type is valid, thereby avoiding power consumption problems caused by unnecessary turning off of the internal combustion engine.

Description

Technical Field

[0001] This application relates to computer technology, and more particularly to a prediction model training method and apparatus, and a parking type prediction method and apparatus. Background Technology

[0002] The start-stop system in a vehicle can automatically control the starting and stopping of the internal combustion engine, thereby reducing the engine's idling time and achieving energy conservation.

[0003] Currently, start-stop systems can shut off the internal combustion engine when they detect the vehicle is stopping and restart it when the stopping action ends. However, because start-stop systems cannot determine the surrounding driving environment or the driver's intentions, they may unnecessarily shut down the vehicle's internal combustion engine, leading to increased fuel consumption. Summary of the Invention

[0004] This application provides a predictive model training method and apparatus, and a parking type prediction method and apparatus, to overcome the problem that the start-stop system shuts down the vehicle's internal combustion engine unnecessarily, thereby increasing the vehicle's energy consumption.

[0005] In a first aspect, embodiments of this application provide a prediction model training method, including:

[0006] The vehicle's driving process is recorded by acquiring a first speed curve and a first relative distance curve, wherein the first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process.

[0007] Based on the first speed curve, determine at least one stopping event during the driving process, and obtain the first stopping position corresponding to the stopping event;

[0008] For any of the parking events, based on the first speed curve and the parking duration threshold, the actual parking type of the parking event is marked as valid or invalid;

[0009] The prediction model is used to process the first speed curve, the first relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

[0010] The model parameters of the prediction model are updated based on the actual parking type and the predicted parking type.

[0011] In one possible design, the process of processing the first speed curve, the first relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model includes:

[0012] Based on the parking time corresponding to the parking event, a portion of the first speed curve within a preset time range before the parking time is determined as the second speed curve;

[0013] A portion of the curves within a preset time range before the parking time in the first relative distance curve is determined as the second relative distance curve;

[0014] The prediction model is used to process the second speed curve, the second relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

[0015] In one possible design, the process of processing the second speed curve, the second relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model includes:

[0016] The second relative distance curve is classified to determine its category;

[0017] Based on the category of the second relative distance curve, determine the influence weight corresponding to the second relative distance curve;

[0018] The prediction model processes the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, to obtain the predicted parking type output by the prediction model.

[0019] In one possible design, the step of processing the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, according to the prediction model to obtain the predicted parking type output by the prediction model includes:

[0020] Feature extraction is performed on the second velocity curve to obtain velocity features;

[0021] The speed feature, the first parking position, and the influence weight are input into the prediction model to obtain the predicted parking type output by the prediction model.

[0022] In one possible design, the classification process for the second relative distance curve to determine its category includes:

[0023] Feature extraction is performed on the second relative distance curve to obtain distance features;

[0024] Unsupervised classification is performed based on the distance features corresponding to each of the multiple parking events, and the multiple distance features are divided into a first category and a second category.

[0025] The category of the second relative distance curve is determined based on the category to which the distance feature of the second relative distance curve belongs.

[0026] In one possible design, the method further includes:

[0027] A distance classification model is trained based on the distance features of the second relative distance curve and the category of the second relative distance curve. The distance classification model is used to output the category based on the distance features.

[0028] In one possible design, labeling the actual parking type of the parking event based on the first speed curve and the parking duration threshold includes:

[0029] Based on the first speed curve, determine the first parking duration of the parking event;

[0030] If the first parking duration is greater than or equal to the parking duration threshold, then the actual parking type is determined to be valid; or,

[0031] If the first parking duration is less than the parking duration threshold, then the actual parking type is determined to be invalid.

[0032] Secondly, embodiments of this application provide a parking type prediction method, including:

[0033] When a vehicle parking event is detected, the target speed curve, target relative distance curve, and target parking position corresponding to the parking event are obtained;

[0034] The target speed curve, the target relative distance curve, and the target parking position are processed by the prediction model to obtain the target parking type output by the prediction model. The target parking type is either valid or invalid. The prediction model is trained according to the method described in the first aspect.

[0035] In one possible design, the process of processing the target speed curve, the target relative distance curve, and the target parking position according to the prediction model to obtain the target parking type output by the prediction model includes:

[0036] Feature extraction is performed on the target relative distance curve to obtain target distance features;

[0037] The target distance feature is input into a distance classification model to obtain the target category output by the distance feature, wherein the target distance feature is trained according to the method described in the first aspect;

[0038] Feature extraction is performed on the target velocity curve to obtain the target velocity features;

[0039] The target speed feature, the target parking location, and the target weight corresponding to the target category are input into the prediction model to obtain the target parking type output by the prediction model.

[0040] In one possible design, the method further includes:

[0041] After the parking event ends, obtain the target parking duration corresponding to the parking event;

[0042] Based on the target parking duration and the parking duration threshold, determine the feedback parking type of the parking event;

[0043] The model parameters of the prediction model are updated based on the feedback parking type and the target parking type.

[0044] In one possible design, the method further includes:

[0045] If the target parking type is valid, then control the vehicle's internal combustion engine to shut down;

[0046] If the target parking type is invalid, the vehicle's internal combustion engine remains running.

[0047] Thirdly, embodiments of this application provide a prediction model training apparatus, comprising:

[0048] The acquisition module is used to acquire a first speed curve and a first relative distance curve corresponding to the vehicle's driving process. The first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process.

[0049] The determining module is configured to determine at least one stopping event during the driving process based on the first speed curve;

[0050] The acquisition module is also used to acquire the first parking location corresponding to the parking event;

[0051] The processing module is used to, for any given parking event, label the actual parking type of the parking event based on the first speed curve and the parking duration threshold, wherein the actual parking type is valid or invalid;

[0052] The processing module is further configured to process the first speed curve, the first relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model.

[0053] The processing module is further configured to update the model parameters of the prediction model based on the actual parking type and the predicted parking type.

[0054] In one possible design, the processing module is specifically used for:

[0055] Based on the parking time corresponding to the parking event, a portion of the first speed curve within a preset time range before the parking time is determined as the second speed curve;

[0056] A portion of the curves within a preset time range before the parking time in the first relative distance curve is determined as the second relative distance curve;

[0057] The prediction model is used to process the second speed curve, the second relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

[0058] In one possible design, the processing module is specifically used for:

[0059] The second relative distance curve is classified to determine its category;

[0060] Based on the category of the second relative distance curve, determine the influence weight corresponding to the second relative distance curve;

[0061] The prediction model processes the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, to obtain the predicted parking type output by the prediction model.

[0062] In one possible design, the processing module is specifically used for:

[0063] Feature extraction is performed on the second velocity curve to obtain velocity features;

[0064] The speed feature, the first parking position, and the influence weight are input into the prediction model to obtain the predicted parking type output by the prediction model.

[0065] In one possible design, the determining module is specifically used for:

[0066] Feature extraction is performed on the second relative distance curve to obtain distance features;

[0067] Unsupervised classification is performed based on the distance features corresponding to each of the multiple parking events, and the multiple distance features are divided into a first category and a second category.

[0068] The category of the second relative distance curve is determined based on the category to which the distance feature of the second relative distance curve belongs.

[0069] In one possible design, the processing module is further configured to:

[0070] A distance classification model is trained based on the distance features of the second relative distance curve and the category of the second relative distance curve. The distance classification model is used to output the category based on the distance features.

[0071] In one possible design, the processing module is specifically used for:

[0072] Based on the first speed curve, determine the first parking duration of the parking event;

[0073] If the first parking duration is greater than or equal to the parking duration threshold, then the actual parking type is determined to be valid; or,

[0074] If the first parking duration is less than the parking duration threshold, then the actual parking type is determined to be invalid.

[0075] Fourthly, embodiments of this application provide a parking type prediction device, comprising:

[0076] The acquisition module is used to acquire the target speed curve, target relative distance curve, and target parking position corresponding to the parking event when a parking event of a vehicle is detected.

[0077] The processing module is used to process the target speed curve, the target relative distance curve, and the target parking position according to the prediction model to obtain the target parking type output by the prediction model, wherein the target parking type is valid or invalid, and the prediction model is trained according to the method described in the first aspect.

[0078] In one possible design, the processing module is specifically used for:

[0079] Feature extraction is performed on the target relative distance curve to obtain target distance features;

[0080] The target distance feature is input into a distance classification model to obtain the target category output by the distance feature, wherein the target distance feature is trained according to the method described in the first aspect;

[0081] Feature extraction is performed on the target velocity curve to obtain the target velocity features;

[0082] The target speed feature, the target parking location, and the target weight corresponding to the target category are input into the prediction model to obtain the target parking type output by the prediction model.

[0083] In one possible design, the processing module is further configured to:

[0084] After the parking event ends, obtain the target parking duration corresponding to the parking event;

[0085] Based on the target parking duration and the parking duration threshold, determine the feedback parking type of the parking event;

[0086] The model parameters of the prediction model are updated based on the feedback parking type and the target parking type.

[0087] In one possible design, the processing module is further configured to:

[0088] If the target parking type is valid, then control the vehicle's internal combustion engine to shut down;

[0089] If the target parking type is invalid, the vehicle's internal combustion engine remains running.

[0090] Fifthly, embodiments of this application provide an electronic device, including:

[0091] Memory, used to store programs;

[0092] A processor for executing the program stored in the memory, wherein, when the program is executed, the processor is configured to perform any of the methods described in the first aspect, the second aspect, various possible designs of the first aspect, and various possible designs of the second aspect.

[0093] In a sixth aspect, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the methods described in the first aspect, the second aspect, various possible designs of the first aspect, and any of various possible designs of the second aspect.

[0094] In a seventh aspect, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the methods described in the first aspect, the second aspect, various possible designs of the first aspect, and any of various possible designs of the second aspect.

[0095] This application provides a prediction model training method and apparatus. The method marks the actual parking type of a vehicle's parking event based on the vehicle's first speed curve, and processes the vehicle's first speed curve, first relative distance curve, and first parking position according to the prediction model to obtain the predicted parking type output by the prediction model. Then, the prediction model is optimized based on the predicted parking type and the actual parking type, thereby training a prediction model that can output accurate predicted parking types.

[0096] Furthermore, this application also provides a parking type prediction method and apparatus. When a parking event is detected, the prediction model can predict whether the parking type of the current parking event is valid or invalid based on the vehicle's speed, the relative distance between the vehicle and the vehicle in front, and the vehicle's parking position. Only when the predicted parking type is valid is the internal combustion engine shut down using the start-stop system. When the predicted parking type is invalid, the start-stop system is not used, and the internal combustion engine remains on. This effectively avoids the high power consumption problem caused by unnecessarily shutting down the internal combustion engine, thereby effectively improving the intelligence and rationality of the start-stop system in shutting down the internal combustion engine. Attached Figure Description

[0097] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0098] Figure 1 A flowchart of the prediction model training method provided in the embodiments of this application;

[0099] Figure 2 A schematic diagram illustrating the implementation of the first velocity curve provided in an embodiment of this application;

[0100] Figure 3 A schematic diagram of the training process of the prediction model provided in the embodiments of this application. Figure 1 ;

[0101] Figure 4 A schematic diagram of the training process of the prediction model provided in the embodiments of this application. Figure 2 ;

[0102] Figure 5 A schematic diagram illustrating the implementation of determining the second velocity curve provided in an embodiment of this application;

[0103] Figure 6 A flowchart of the parking type prediction method provided in the embodiments of this application;

[0104] Figure 7 A schematic diagram illustrating the processing procedure of the prediction model provided in the embodiments of this application;

[0105] Figure 8 A schematic diagram illustrating the implementation of the optimized prediction model provided in the embodiments of this application;

[0106] Figure 9 This is a schematic diagram of the structure of the prediction model training device provided in the embodiments of this application;

[0107] Figure 10 This is a schematic diagram of the parking type prediction device provided in the embodiments of this application;

[0108] Figure 11 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0109] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0110] Before introducing this application, it should be noted that the technical solution provided in this application is particularly applicable to commercial vehicles, where commercial vehicles refer to automobiles designed and technically characterized for transporting people and goods. For example, a commercial vehicle may include a freight vehicle and a bus with more than n seats, where n may have different settings in different countries and regions, and this embodiment does not impose any limitations on this.

[0111] However, it should be understood that this does not mean that the technical solution of this application can only be applied to commercial vehicles. Any vehicle that includes an internal combustion engine can adopt the technical solution of this application to achieve the purpose of energy saving. This embodiment does not limit the specific application vehicle.

[0112] Based on this, in order to better understand the technical solution of this application, the relevant technologies involved in this application will be further described in detail below.

[0113] A start-stop system can be installed in a vehicle. The function of the start-stop system is to automatically shut down and restart the internal combustion engine to reduce the engine's idling time, thereby effectively reducing the vehicle's fuel consumption and emissions.

[0114] In one implementation, the start-stop system can shut down the internal combustion engine when it detects that the vehicle has stopped. Alternatively, the start-stop system can also shut down the internal combustion engine by combining certain auxiliary conditions when it detects that the vehicle has stopped. These auxiliary conditions could be, for example, that the vehicle's current fuel level is less than a fuel level threshold, or that the battery charge is less than a battery charge threshold, etc.

[0115] Furthermore, after the start-stop system shuts down the internal combustion engine, it can continue to detect whether the parking behavior has ended. If it determines that the parking behavior of the vehicle has ended, it can further control the internal combustion engine to shut down, thereby automatically restoring the vehicle to its starting state.

[0116] However, traditional start-stop systems cannot determine the driving environment around the vehicle or the driver's parking intentions, which may cause the vehicle's internal combustion engine to shut down unnecessarily.

[0117] For example, during peak hours, a vehicle's idling time may be very short (idling refers to the engine running at idle speed). In such scenarios, drivers may frequently make very short stops due to traffic jams, such as stopping the vehicle for 1 to 2 seconds and then quickly starting it again.

[0118] If the start-stop system controls the starting and stopping of the internal combustion engine based on the aforementioned technical solution, it may lead to frequent engine shutdowns and restarts. Each start-up consumes electrical energy, so excessively frequent engine starts can actually result in poor vehicle fuel efficiency. Furthermore, the vehicle starts much faster when the engine is off and then starting up compared to when it's on. This delayed start-up leads to a poor driving experience for the driver, and the engine's shut-off and restarting causes vibrations. Frequent unnecessary vibrations during start-up can negatively impact passenger comfort.

[0119] Based on the problems described above, this application proposes the following technical concept: It is necessary to predict each parking action, specifically whether the parking action is valid or invalid. The start-stop system intervenes only for parking actions predicted as valid, i.e., shutting off the internal combustion engine. For parking actions predicted as invalid, the start-stop system does not intervene, keeping the internal combustion engine running, thereby avoiding unnecessary engine shutdown and specifically addressing the aforementioned technical problems.

[0120] When devising a technical solution for determining the validity of parking behavior, the following options are available:

[0121] Option 1: By defining preset rules and judging whether the parking data in the last n seconds meets the preset rules, the validity of the parking behavior can be estimated.

[0122] However, different drivers have different driving habits, so it is impossible to define a set of universal preset rules to cover all scenarios, which may lead to incorrect predictions of the effectiveness of parking behavior.

[0123] Option 2: Receive motion status information of other vehicles on the road via wireless communication to estimate the likelihood of the vehicle stopping. However, wireless communication between vehicles is currently not easy to achieve, therefore this approach is not applicable.

[0124] Option 3: Obtain vehicle driving environment data, route data, traffic data on the road, etc., and predict the effectiveness of vehicle parking based on data from these different sources.

[0125] However, the data from these different sources mentioned above are not easy to obtain, so this implementation method is not applicable.

[0126] Given the shortcomings of the aforementioned solutions, this application proposes a prediction model that can estimate the effectiveness of a vehicle's parking behavior based on its own data. The training and application processes of the prediction model are described below.

[0127] First, the training process of the prediction model is explained. The execution subject of the prediction model training method can be a local server, a cloud server, a processor, a chip, or other devices with data processing capabilities. This embodiment does not limit the specific execution subject, and it can be selected and set according to actual needs.

[0128] First, combine Figures 1 to 3 This paper introduces the training methods for prediction models. Figure 1 This is a flowchart of the prediction model training method provided in the embodiments of this application. Figure 2 This is a schematic diagram illustrating the implementation of the first velocity curve provided in an embodiment of this application. Figure 3 A schematic diagram of the training process of the prediction model provided in the embodiments of this application. Figure 1 .

[0129] like Figure 1 As shown, the method includes:

[0130] S101. Obtain the first speed curve and the first relative distance curve corresponding to the vehicle's driving process. The first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process.

[0131] In this embodiment, when training the prediction model, the driving data during the vehicle's journey can be used as the training data for the prediction model. Specifically, the process from when the vehicle starts from a first parking state until it returns to the first parking state can be understood as one driving cycle. The first parking state can be understood, for example, as the vehicle's parking time exceeding a set duration. The first parking state can be understood as the vehicle ending its journey, rather than a brief stop.

[0132] For example, the driving data during the vehicle's operation may include a speed curve and a relative distance curve. In this embodiment, the speed curve used for training the prediction model is referred to as the first speed curve, and the relative distance curve used for training the prediction model is referred to as the first relative distance curve.

[0133] The first speed curve indicates the vehicle's speed at various moments during its journey, for example, it can be referenced... Figure 2 To understand the first velocity curve, such as... Figure 2 As shown, the horizontal axis of the first velocity curve represents time t, and the vertical axis represents the vehicle's speed v.

[0134] For example in Figure 2 In the example, suppose the vehicle starts moving from its initial stationary state at time a, and its speed at time a is 0 km / h. After time a, the vehicle begins to accelerate until time b, where it is assumed that the vehicle's speed becomes 45 km / h.

[0135] Then, from time b to time c, assume the vehicle maintains a constant speed of 45 km / h. And in... Figure 2 In the example, after time c, the vehicle's speed begins to decrease. Assuming that by time d, the vehicle's speed drops to 0 km / h, this means that the vehicle stops at time d.

[0136] The above is based on Figure 2 Taking time a to time d as an example, for Figure 2 The vehicle's driving behavior, characterized by the first velocity curve, is illustrated in the figure. Figure 2 After time d, the first velocity curve represents the vehicle's driving behavior in a similar way, which will not be repeated here.

[0137] The above combination Figure 2 An example of a first speed curve is presented. The first relative distance curve is similar to the first speed curve, except that the first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front at various times during the driving process.

[0138] It should be noted that the "front vehicle" mentioned in the various embodiments of this application does not refer to a specific vehicle. It can be understood as the vehicle located directly in front of the vehicle during the vehicle's operation. In other words, the "front vehicle" is a relative concept, and the vehicle it refers to can change according to actual needs.

[0139] For example, in the first relative distance curve, the horizontal axis represents time t, and the vertical axis represents the relative distance d between the vehicle and the vehicle in front. For instance, assuming that the relative distance at time a is 3 meters in the first relative distance curve, it means that at time a, the relative distance between the vehicle and the vehicle in front is 3 meters. The remaining specific implementations of the first relative distance curve in this embodiment will not be elaborated further; they can be referred to the first speed curve described above.

[0140] S102. Based on the first speed curve, determine at least one stopping event during the driving process, and obtain the first stopping position corresponding to the stopping event.

[0141] In this embodiment, the first speed curve can indicate the vehicle's speed at various times during the driving process. When the vehicle's speed is 0, it can be determined that the vehicle has stopped. In this embodiment, a single stopping behavior of the vehicle can be understood as a stopping event.

[0142] Therefore, as Figure 3 As shown, at least one stopping event during the driving process can be determined based on the first speed curve. For example, in the above... Figure 2 In the example, the vehicle's speed is 0 at times a, b, e, and f, so it can be determined that a parking event occurred at these four times, thus identifying four parking events.

[0143] Furthermore, in this embodiment, the first parking position corresponding to the parking event can also be obtained. During the vehicle's driving process, the vehicle's driving data can also include the vehicle's position at various times. So, after determining the parking time with a speed of 0 as described above, the position of the vehicle at the parking time can be determined as the first parking position corresponding to the parking event.

[0144] The parking location can be a GPS (Global Positioning System) location, or any other possible form, as long as it can represent the vehicle's location.

[0145] For example, as described above Figure 2 In the above, assuming the vehicle is at position one at time a, then the first parking position corresponding to the parking event occurring at time a is position one. Similarly, assuming the vehicle is at position two at time d, then the first parking position corresponding to the parking event occurring at time d is position two.

[0146] S103. For any parking event, based on the first speed curve and the parking duration threshold, mark the actual parking type of the parking event, which is either valid or invalid.

[0147] To avoid the problem of high vehicle energy consumption caused by the start-stop system frequently shutting down the internal combustion engine in response to parking events, one approach is to predict whether the current parking behavior is effective.

[0148] Specifically, when a parking event is valid, the start-stop system can respond to the event and shut down the internal combustion engine to save fuel. Conversely, when a parking event is invalid, the system can disregard the event and keep the internal combustion engine running, thus avoiding the higher energy consumption caused by frequent engine switching, as mentioned above.

[0149] In this embodiment, the purpose of training the prediction model is to predict the effectiveness of parking behavior, thereby providing a basis for subsequent processing regarding whether to shut down the internal combustion engine based on the start-stop system. During the model training phase, the training data needs to be labeled to clarify what specific parking behaviors are effective and what are ineffective, so that the prediction model can learn the corresponding features and achieve the goal of predicting the effectiveness of parking behavior.

[0150] like Figure 3 As shown, for example, for any parking event, its actual parking type can be labeled, where the parking type is valid or invalid.

[0151] In one possible implementation, the duration of parking can be used as a criterion to distinguish between valid and invalid parking. For example, a parking duration threshold can be set. When the parking duration corresponding to a parking event is greater than or equal to the threshold, the parking event can be determined to be valid. Conversely, when the parking duration corresponding to a parking event is less than the threshold, the parking event can be determined to be invalid.

[0152] For example, the parking duration threshold can be represented by k, where k is a value greater than or equal to 0. For instance, when k is set to 8, and assuming the unit of the parking duration threshold is seconds, then if the parking duration corresponding to a parking event is greater than or equal to 8 seconds, the parking event can be determined to be valid. Alternatively, if the parking duration corresponding to a parking event is less than 8 seconds, the parking event can be determined to be invalid.

[0153] In practical implementation, a parking time threshold can be preset for each vehicle type. The parking time threshold can vary between different vehicle types, which can include at least one of the following: sedans, SUVs, trucks, etc. For example, the parking time threshold for different vehicle types can range from 6 to 12 seconds.

[0154] Alternatively, in addition to being related to the vehicle model, the parking time threshold can be adaptively adjusted according to each driver's wishes, and this embodiment does not limit the specific implementation method of the parking time threshold.

[0155] S104. Based on the prediction model, process the first speed curve, the first relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

[0156] After preparing the training data, such as Figure 3 As shown, in this embodiment, the first speed curve, the first relative distance curve, and the first parking position can be processed according to the prediction model. After processing, the predicted parking type output by the prediction model can be obtained, where the predicted parking type is valid or invalid.

[0157] It is understood that the prediction model in this embodiment refers to three aspects of data when predicting whether the parking type of the vehicle is valid or invalid: the vehicle's driving speed reflected by the first speed curve, the relative distance between the vehicle and the vehicle in front reflected by the first relative distance curve, and the parking position of the vehicle reflected by the first parking position.

[0158] First, let's introduce the first velocity curve:

[0159] Specifically, different drivers may have different driving habits, and each user's speed changes before parking may also exhibit certain characteristics. For example, a user who habitually maintains a relatively high vehicle speed on the road will reduce the vehicle's speed from a relatively high value to 0 when they need to perform the effective parking described above, thus bringing the vehicle to a stop.

[0160] In the peak-hour scenarios described above, due to road conditions, the user may only be able to maintain a relatively slow speed to control the vehicle. Therefore, when parking, the vehicle speed must be reduced from a relatively low value to zero to bring it to a stop. Thus, for the same driver, the characteristics of the vehicle's speed will differ depending on whether the parking is effective or ineffective. Therefore, the first speed curve can, to some extent, reflect the effectiveness or ineffectiveness of a parking action.

[0161] Furthermore, the time it takes for a user to reduce their speed to zero may also exhibit certain characteristics. For example, some users habitually brake slowly, maintaining a longer braking time to achieve a gradual stop. However, in the peak-hour scenarios described above, road conditions limit the ability for users to brake slowly. Braking time can also be reflected in the first speed curve, thus allowing the first speed curve to indicate the effectiveness or ineffectiveness of a stopping action from multiple perspectives.

[0162] Next, let's talk about the first parking location:

[0163] For each driver, their parking locations also exhibit certain characteristics. For example, a driver who frequently commutes between their workplace and residential area might often park in or near their workplace. Similarly, they might frequently park in or near their residential area; these are the locations where their vehicle is typically parked.

[0164] Furthermore, for this driver, who may always commute along the same route, the traffic light intersections on that route, or the locations within a predetermined range of traffic light intersections, may also be the places where the vehicle frequently stops.

[0165] Therefore, if the first parking location corresponding to the parking behavior is one of the locations mentioned above, it is highly likely to be a valid parking instance. Otherwise, it may be due to some special circumstances that the vehicle is parked in an infrequently used parking location, in which case it may be an invalid parking instance.

[0166] Therefore, the first parking position can also reflect the effectiveness or ineffectiveness of a parking behavior to some extent.

[0167] Finally, let's discuss the first relative distance curve:

[0168] When a vehicle is traveling on the road, if the relative distance between the vehicle and the vehicle in front is relatively close, the impact of the vehicle in front on the vehicle (which can also be understood as the vehicle itself) is usually relatively large. For example, when the vehicle in front stops, the stopping behavior of the vehicle in front will be affected to some extent by the stopping behavior of the vehicle in front. Suppose the vehicle in front brakes suddenly to a stop, then the vehicle in front may also brake suddenly to a stop in order to avoid a collision.

[0169] When the relative distance between a vehicle and the vehicle in front is relatively large, the impact of the vehicle in front on the vehicle is relatively small. In this case, the vehicle's parking behavior is likely to be in line with the driver's driving habits.

[0170] Therefore, the first relative distance curve in this embodiment can reflect the degree of influence of the preceding vehicle on the current vehicle when the parking behavior occurs.

[0171] Meanwhile, the three parameters mentioned above are all data of the vehicle itself, which are relatively easy to obtain. Therefore, training the prediction model based on these three types of data can effectively improve the adaptability of the technical solution of this application.

[0172] It should also be noted that the training data collected in this application during the training of the prediction model is all historical driving data of the same vehicle. In other words, this application trains a prediction model specifically for each vehicle. Since the three driving parameters described above are all historical driving data of the vehicle itself, a prediction model can be built based on the vehicle's historical data. This allows the prediction model to specifically learn the parking characteristics of the vehicle (or the driver), ensuring that each prediction model in this application is personalized for the vehicle (or the driver), thereby improving the accuracy of the subsequent output prediction of the parking type.

[0173] S105. Update the model parameters of the prediction model based on the actual parking type and the predicted parking type.

[0174] During the model training phase, the accuracy of the predicted parking type output by the prediction model cannot be guaranteed. In order to improve the output accuracy of the prediction model, the model parameters of the prediction model can be updated based on the accurate real parking type obtained from the standard and the predicted parking type output by the prediction model, thereby optimizing the prediction model so that the predicted parking type output by the prediction model can be close to the accurate real parking type.

[0175] In one possible implementation, a first loss value can be determined based on the actual parking type and the predicted parking type. Then, the model parameters of the prediction model are updated based on this first loss value, thereby optimizing the prediction model so that the predicted parking type output by the model is as close as possible to the accurate actual parking type. This ensures that the trained prediction model can output an accurate predicted parking type during model usage.

[0176] For example, supervised classification algorithms such as decision trees or SVM (support vector machines) can be used to generate prediction models. The specific generation method or network structure of the prediction model can also be selected and set according to actual needs.

[0177] The prediction model training method provided in this application includes: acquiring a first speed curve and a first relative distance curve corresponding to the vehicle's driving process, wherein the first relative distance curve indicates the relative distance between the vehicle and the vehicle in front during the driving process. Based on the first speed curve, at least one parking event during the driving process is determined, and a first parking position corresponding to the parking event is obtained. For any parking event, based on the first speed curve and a parking duration threshold, the actual parking type of the parking event is labeled, wherein the actual parking type is valid or invalid. The first speed curve, the first relative distance curve, and the first parking position are processed according to the prediction model to obtain the predicted parking type output by the prediction model. The model parameters of the prediction model are updated based on the actual parking type and the predicted parking type. By labeling the actual parking type of the vehicle's parking events based on the vehicle's first speed curve, and processing the vehicle's first speed curve, the first relative distance curve, and the first parking position according to the prediction model, the predicted parking type output by the prediction model is obtained. Then, the prediction model is optimized based on the predicted parking type and the actual parking type, thereby training a prediction model capable of outputting accurate predicted parking types.

[0178] Furthermore, during training, the first speed curve, the first relative distance curve, and the first parking position can all reflect parking characteristics to a certain extent. Therefore, based on the first speed curve, the first relative distance curve, and the first parking position, sufficient information can be provided to the prediction model so that it can output the predicted parking type. Also, the information described above is all vehicle-specific information, so its acquisition difficulty is relatively low. Therefore, training the prediction model has good adaptability and feasibility.

[0179] Based on the above introduction, the following will combine... Figures 4 to 5 In the prediction model training method provided in this application, the prediction model processes the first speed curve, the first relative distance curve, and the first parking type to output the possible implementation process of predicting the parking type, which will be further described in detail later. Figure 4 A schematic diagram of the training process of the prediction model provided in the embodiments of this application. Figure 2 , Figure 5 This is a schematic diagram illustrating the implementation of determining the second velocity curve in an embodiment of this application.

[0180] In one possible implementation, before processing the first speed curve, the first relative distance curve, and the parking position according to the preset curve, these three data can be preprocessed first.

[0181] For example, you can refer to Figure 4For example, the second velocity curve can be determined based on the first velocity curve, and the second relative distance curve can be determined based on the first relative distance curve.

[0182] In this embodiment, the first speed curve and the first relative distance curve are both complete curves corresponding to one driving process of the vehicle. In order to improve the effectiveness and specificity of subsequent feature extraction, for example, a portion of the curves in the first speed curve and the first relative distance curve that occurs a certain period of time before the stopping behavior occurs can be extracted to obtain the second speed curve and the second relative distance curve.

[0183] In one possible implementation, the parking time corresponding to the parking event can be determined, and then a portion of the first speed curve within a preset time range before the parking time can be determined as the second speed curve.

[0184] For example, you can refer to Figure 5 To understand, Figure 5 The upper half of the curve represents the first velocity curve, where the horizontal axis represents time and the vertical axis represents vehicle speed. (Refer to...) Figure 5 The schematic first velocity curve shows that the vehicle stopped at times a, d, e, and f.

[0185] For example, taking parking behavior at time d as an example, we can determine that the parking time corresponding to parking behavior at time d is time d itself, and assume that the preset duration is set to T1. Then, as... Figure 5 As shown, the time corresponding to the preset duration T1 before the stopping time d is time g. Therefore, the portion of the first velocity curve from time g to time d can be defined as follows: Figure 5 The second velocity curve is shown in the lower half of the image.

[0186] Similar to the processing method for the first relative curve, a portion of the first relative distance curve within a preset time range before the stopping time can be determined as the second relative distance curve. The specific implementation method can be referred to the processing method for the first speed curve described above, and will not be repeated here.

[0187] In actual implementation, the preset duration set for the first velocity curve and the preset duration set for the first relative distance curve can be equal or unequal; this embodiment does not impose any restrictions on this. The specific implementation of the preset duration can be selected according to actual needs.

[0188] In this embodiment, the second speed curve and the second relative distance curve can specifically reflect the vehicle's speed and relative distance to the vehicle in front within a preset time range before the parking event occurs. Therefore, the second speed curve and the second relative distance curve are then processed according to the prediction model, which can effectively improve the correlation between the extracted features and the parking behavior, reduce the amount of subsequent data processing, and improve the model training efficiency.

[0189] Next refer to Figure 4 The subsequent processing after obtaining the second velocity curve and the second relative distance curve is explained.

[0190] like Figure 4 As shown, in this embodiment, feature extraction can be performed on the second relative distance curve to obtain distance features. In one implementation, this embodiment can determine the average value of the relative distance and / or the rate of change of the relative distance based on the second relative distance curve, wherein the distance features may include, for example, the average value of the relative distance and / or the rate of change of the relative distance described herein. Alternatively, the distance features may further include the maximum value, minimum value, etc., of the distance. This embodiment does not limit the specific content of the distance features; any data obtained from the second relative distance curve that reflects the characteristics of the relative distance can be used as the distance features in this embodiment.

[0191] The data format for distance features can be, for example, a distance feature vector, or any other possible data format.

[0192] Furthermore, in this embodiment, feature extraction can be performed on the second velocity curve to obtain velocity features. In one implementation, this embodiment can determine at least one of the following data based on the second velocity curve: velocity value, average velocity, maximum velocity, minimum velocity, average acceleration, maximum acceleration, average deceleration, and maximum deceleration, etc., wherein the velocity features may include, for example, the various velocity data described above. Alternatively, the velocity features may further include the minimum acceleration, minimum deceleration, etc. This embodiment does not limit the specific content of the velocity features; any data obtained from the second velocity curve that reflects velocity characteristics can be used as the velocity features in this embodiment.

[0193] The data format for velocity features can be, for example, a velocity feature vector, or any other possible data format.

[0194] Once the velocity and distance features are determined, the system is ready to input these features into the prediction model so that the model can process the data.

[0195] In one possible implementation, the technical solution of this application can directly normalize the distance features, speed features, and the first parking position described above, and then input the normalized feature vector into the prediction model to obtain the predicted parking type output by the prediction model.

[0196] Alternatively, in another possible implementation, this embodiment can further determine the influence weight based on distance features. The influence weight indicates the degree of influence of the preceding vehicle on the vehicle. Then, the influence weight, speed features, and the first parking position are processed according to the prediction model to obtain the predicted parking type.

[0197] Here is a brief introduction to how the influence weights are determined:

[0198] Based on the above embodiments, it can be determined that the first relative distance curve in this embodiment can reflect the degree of influence of the preceding vehicle on the current vehicle when parking occurs, and the selected second relative distance curve also has the same function.

[0199] Furthermore, during the model training phase, a large amount of training data is available. For example, for the same vehicle, the first speed curve, first relative distance curve, and first parking position can be extracted for each of the multiple driving processes. Assuming that the technical solution described above is executed for each driving process, the distance features corresponding to multiple parking events in multiple driving processes can be obtained.

[0200] Subsequently, for example, classification can be performed based on the distance features corresponding to each of the multiple parking events, thereby dividing the moment features corresponding to each of the multiple parking events into a first category and a second category. The classification process can employ algorithms such as DBSCAN (density-based spatial clustering of applications with noise) or k-means for unsupervised classification, or a binary classification algorithm. This embodiment does not limit the specific classification method.

[0201] Understandably, classification processes group distance features with similar characteristics into the same category. Therefore, distance features in the first category are relatively similar, as are those in the second category. In this embodiment, the distance features reflect the relative distance between the vehicle and the vehicle in front for a period of time before the parking action occurs.

[0202] In one possible implementation, among the first and second categories, the distance feature in one category reflects the relative distance between the vehicle and the vehicle in front during a period of time before the parking action occurred. This means the vehicle in front has a significant influence on the vehicle's parking behavior. Therefore, a smaller weight can be assigned to this category to weaken the expressive power of speed and parking position features when the vehicle in front has a significant influence on the vehicle's parking behavior, thus avoiding adverse interference to the prediction model's learning of vehicle parking behavior characteristics. The reason for this interference is that the current parking behavior characteristics may not necessarily conform to the driver's parking habits and may be influenced by the vehicle in front.

[0203] Furthermore, within the first and second categories, the distance feature in the other category reflects the relative distance between the vehicle and the vehicle in front for a period of time before the parking action occurs. This means the influence of the vehicle in front on the vehicle's parking behavior is relatively small. Therefore, a larger weight can be assigned to this category to ensure or enhance the expressive power of speed and parking position features when the influence of the vehicle in front is minimal. This allows the prediction model to learn more accurate parking behavior characteristics. This is because the current parking behavior is less affected by the vehicle in front, and therefore is likely consistent with the driver's parking habits.

[0204] After completing the classification described above, the category corresponding to the second relative distance curve is determined based on the category to which the distance feature of the currently processed second relative distance curve belongs. Then, the influence weight corresponding to the second relative distance curve is determined based on the category to which the second relative distance curve belongs.

[0205] For example, assuming that the first category described above is the category with a greater degree of influence, and the second category is the category with a smaller degree of influence, then the weight of the first category is less than the weight of the second category.

[0206] If the distance feature of the second relative distance curve belongs to the first category described above, then the weight corresponding to the first category can be determined as the influence weight corresponding to the second relative distance curve. Alternatively, if the distance feature of the second relative distance curve belongs to the second category described above, then the weight corresponding to the second category can be determined as the influence weight corresponding to the second relative distance curve.

[0207] After obtaining the data described above, refer to Figure 4 The speed characteristics, the first parking position, and the influence weights can be input into the prediction model to obtain the predicted parking type output by the prediction model.

[0208] Optionally, before inputting the prediction model, the speed features and the first parking position can be normalized. Then, the normalized features and influence weights are input into the prediction model to obtain the predicted parking type output by the prediction model.

[0209] In this embodiment, by performing the aforementioned processing on the first speed curve, the first relative distance curve, and the first parking position, the influence weights, speed characteristics, and the first parking position that can be input into the prediction model can be effectively obtained. These data are then input into the prediction model so that the prediction model can complete the corresponding feature analysis and data processing, thereby effectively outputting the predicted parking type.

[0210] Furthermore, during the model training phase, due to the large amount of data, the classification method described above can be used to determine the category of distance features, thereby determining the influence weight of distance features. However, in practical applications, a single parking behavior may only have a single distance feature, making it impossible to use the classification method described here to determine the category of the distance feature.

[0211] Therefore, in one possible implementation, this embodiment can also use the distance features corresponding to the multiple parking events described above as training data, and the classification results for the multiple distance features as labeled data (i.e., accurate classification results) to train a distance classification model. The input of the distance classification model is the distance feature, and the output of the distance classification model is the category to which the distance feature belongs. The training method of the distance classification model is similar to the training method of the prediction model described above, and will not be repeated here.

[0212] The above embodiments describe the training method of the prediction model. After the prediction model is trained, it can be used to predict the parking type of parking behavior. The following describes the usage process of the prediction model in conjunction with specific embodiments, that is, the parking type prediction method provided in this application.

[0213] For example, the execution entity of the parking type prediction method in this application can be a vehicle, or it can be a processor or chip in the vehicle, etc. It is understood that the trained prediction model can be deployed in the vehicle, which can collect driving data during operation. When parking is detected, the driving data can be processed according to the prediction model deployed in the vehicle to output a predicted parking type for the current parking event. Then, based on the predicted parking type, a decision is made as to whether to shut down the internal combustion engine through the start-stop system.

[0214] The following is a combination of... Figure 6 This application introduces the parking type prediction method. Figure 6A flowchart of the parking type prediction method provided in the embodiments of this application.

[0215] like Figure 6 As shown, the method includes:

[0216] S601. When a vehicle parking event is detected, acquire the target speed curve, target relative distance curve, and target parking position corresponding to the parking event.

[0217] In one possible implementation, when a vehicle's speed is detected to be 0, a parking event can be determined. Upon determining that a parking event has occurred, the technical solution of this embodiment is executed to predict whether the current parking event is valid or invalid.

[0218] In this embodiment, in order to provide data to the prediction model, when a vehicle parking event is detected, the target speed curve, target relative distance curve, and target parking position corresponding to the parking event can be obtained first.

[0219] The target speed curve can be the curve corresponding to the vehicle's speed during this driving process. Alternatively, the target speed curve can also be the curve corresponding to the vehicle's speed within a preset time range before the stopping time corresponding to the current stopping event.

[0220] Furthermore, the target relative distance curve is the curve corresponding to the relative distance between the vehicle and the vehicle in front during this driving process. Alternatively, the target relative distance can also be the curve corresponding to the relative distance between the vehicle and the vehicle in front within a preset time range before the stopping time corresponding to the current stopping event.

[0221] In addition, the target parking location is the location of the vehicle at the time the current parking event occurs.

[0222] S602. Based on the prediction model, process the target speed curve, target relative distance curve, and target parking position to obtain the target parking type output by the prediction model, which is either valid or invalid.

[0223] After obtaining the target speed curve, target relative distance, and target parking position as described above, the target speed curve, target relative distance, and target parking position can be processed according to the prediction model to obtain the target parking type output by the prediction model. The target parking type is the parking type predicted for the current parking event, which is either valid or invalid.

[0224] The prediction model is processed in a similar way to the training process described above. The prediction model in this embodiment is trained based on the methods described in the above embodiments. Therefore, the prediction model can effectively predict the parking type of parking events, thereby ensuring the accuracy of the output target parking type to a certain extent.

[0225] In one implementation, if the target parking type is valid, it can be predicted that the parking duration corresponding to the current parking event is relatively long. Therefore, the internal combustion engine of the vehicle can be shut down based on the start-stop system, thereby saving fuel consumption.

[0226] In another implementation, if the target parking type is invalid, the parking duration corresponding to the current parking event can be predicted to be short. Therefore, the start-stop system can be omitted. In other words, even if a parking event occurs, the vehicle's internal combustion engine remains running. This effectively avoids the high power consumption problem caused by the frequent switching of the internal combustion engine, as described above.

[0227] In the parking type prediction method provided in this application, when a parking event is detected, the prediction model can predict whether the parking type of the current parking event is valid or invalid based on the vehicle's speed, the relative distance between the vehicle and the vehicle in front, and the vehicle's parking position. Only when the predicted parking type is valid is the internal combustion engine shut down using the start-stop system. When the predicted parking type is invalid, the start-stop system is not used, and the internal combustion engine remains on. This effectively avoids the problem of high power consumption caused by unnecessarily shutting down the internal combustion engine, thereby effectively improving the intelligence and rationality of the start-stop system in shutting down the internal combustion engine.

[0228] Based on the above introduction, the following will further combine... Figure 7 This section provides a more detailed introduction to the specific data processing methods used in the prediction model. Figure 7 This is a schematic diagram illustrating the processing procedure of the prediction model provided in the embodiments of this application.

[0229] like Figure 7 As shown, whenever a new parking event occurs, the vehicle can obtain the target relative distance curve, target speed curve, and target parking position corresponding to the parking event.

[0230] In one possible implementation, feature extraction can be performed on the target relative distance curve to obtain the target distance feature. The target distance feature here is similar to the distance feature introduced in the model training phase above, and will not be repeated here.

[0231] After obtaining the target distance features, such as Figure 7As shown, for example, the target distance feature can be input into the trained distance classification model, and the distance classification model can output the target category corresponding to the target distance feature, where the target category is either the first category or the second category described above.

[0232] and reference Figure 7 In this embodiment, the target weights corresponding to the target categories can also be obtained, awaiting subsequent input into the prediction model. The implementation of the target weights is similar to that of the influence weights described above, and will not be repeated here.

[0233] and reference Figure 7 In this embodiment, feature extraction can also be performed on the target velocity curve to obtain the target velocity features. The target velocity features here are similar to the velocity features introduced in the model training stage above, and will not be described again here.

[0234] After obtaining the target velocity characteristics, such as Figure 7 As shown, for example, normalization can be performed on the target speed characteristics and the target parking position.

[0235] Then, the target weights, normalized target speed features, and target parking position can be input into the prediction model to obtain the target parking details output by the prediction model.

[0236] The processing details of the prediction model during the usage phase are similar to those of the processing details of the prediction model during the training process described above. Details not described in this embodiment can be understood by referring to the content described in the above embodiments.

[0237] Based on the above embodiments, it can be understood that the target parking type is the parking type predicted by the prediction model. Although the prediction model has undergone a training process and can guarantee the accuracy of the output results to a certain extent, the target parking type output by the prediction model can still be verified.

[0238] Specifically, the validity or invalidity indicated by the target parking type mainly corresponds to the length of time the vehicle is parked, which can be measured using the parking duration threshold mentioned above.

[0239] For example, when the target parking type is valid, it means that the predicted parking time of the vehicle will be greater than or equal to the parking time threshold.

[0240] For example, if the target parking type is invalid, it means that the predicted parking time of the vehicle will be less than the parking time threshold.

[0241] The above describes a predicted parking duration. Once the parking event ends and the vehicle starts moving, the actual parking duration can be determined, thus verifying the accuracy of the predicted target parking type. Subsequently, the prediction model parameters can be updated based on the target parking type and the actual parking type to continuously optimize the prediction model during its use.

[0242] In one implementation, the target parking duration corresponding to the parking event can be obtained after the parking event ends. For example, when the vehicle's speed is greater than 0, the parking time can be considered to have ended.

[0243] Then, the feedback parking type of the parking event can be determined based on the target parking duration and the parking duration threshold. For example, when the target parking duration is greater than or equal to the parking duration threshold, the feedback parking type of the parking event is determined to be valid; and when the target parking duration is less than the parking duration threshold, the feedback parking type of the parking event is determined to be invalid.

[0244] Then refer to Figure 8 The model parameters of the prediction model can be updated based on the feedback parking type and the target parking type. Figure 8 This is a schematic diagram illustrating the implementation of the optimized prediction model provided in the embodiments of this application.

[0245] For example, a second loss value can be determined based on the feedback parking type and the target parking type, and then the model parameters of the prediction model can be updated based on the second loss value, so that the difference between the target parking type output by the prediction model and the feedback parking type can become smaller and smaller, thereby improving the accuracy of the target parking type output by the prediction model.

[0246] Furthermore, when updating the model parameters of the prediction model, the model input can also be referenced, for example, Figure 7 As shown, in addition to the target parking type and the feedback parking type, the target influence weight and normalized features, which are used as inputs to the model, can also be referenced. The normalized features are the target speed features and the target parking position after normalization. Then, the model parameters of the prediction model are updated based on these data to obtain the optimized prediction model. The implementation method is similar to that described above. It can also be to determine the loss value and then optimize the model parameters based on the loss value.

[0247] Therefore, the prediction model provided in this application can be continuously and dynamically optimized during use to improve the accuracy of the vehicle parking type predicted by the prediction model.

[0248] Figure 9 This is a schematic diagram of the predictive model training device provided in an embodiment of this application. Figure 9 As shown, the device 90 includes: an acquisition module 901, a determination module 902, and a processing module 903.

[0249] The acquisition module 901 is used to acquire a first speed curve and a first relative distance curve corresponding to the driving process of the vehicle. The first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process.

[0250] The determining module 902 is used to determine at least one stopping event during the driving process based on the first speed curve;

[0251] The acquisition module 901 is further configured to acquire the first parking location corresponding to the parking event;

[0252] Processing module 903 is used to, for any one of the parking events, mark the actual parking type of the parking event according to the first speed curve and the parking duration threshold, wherein the actual parking type is valid or invalid;

[0253] The processing module 903 is further configured to process the first speed curve, the first relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model.

[0254] The processing module 903 is further configured to update the model parameters of the prediction model based on the actual parking type and the predicted parking type.

[0255] In one possible design, the processing module 903 is specifically used for:

[0256] Based on the parking time corresponding to the parking event, a portion of the first speed curve within a preset time range before the parking time is determined as the second speed curve;

[0257] A portion of the curves within a preset time range before the parking time in the first relative distance curve is determined as the second relative distance curve;

[0258] The prediction model is used to process the second speed curve, the second relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

[0259] In one possible design, the processing module 903 is specifically used for:

[0260] The second relative distance curve is classified to determine its category;

[0261] Based on the category of the second relative distance curve, determine the influence weight corresponding to the second relative distance curve;

[0262] The prediction model processes the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, to obtain the predicted parking type output by the prediction model.

[0263] In one possible design, the processing module 903 is specifically used for:

[0264] Feature extraction is performed on the second velocity curve to obtain velocity features;

[0265] The speed feature, the first parking position, and the influence weight are input into the prediction model to obtain the predicted parking type output by the prediction model.

[0266] In one possible design, the determining module 902 is specifically used for:

[0267] Feature extraction is performed on the second relative distance curve to obtain distance features;

[0268] Unsupervised classification is performed based on the distance features corresponding to each of the multiple parking events, and the multiple distance features are divided into a first category and a second category.

[0269] The category of the second relative distance curve is determined based on the category to which the distance feature of the second relative distance curve belongs.

[0270] In one possible design, the processing module 903 is further configured to:

[0271] A distance classification model is trained based on the distance features of the second relative distance curve and the category of the second relative distance curve. The distance classification model is used to output the category based on the distance features.

[0272] In one possible design, the processing module 903 is specifically used for:

[0273] Based on the first speed curve, determine the first parking duration of the parking event;

[0274] If the first parking duration is greater than or equal to the parking duration threshold, then the actual parking type is determined to be valid; or,

[0275] If the first parking duration is less than the parking duration threshold, then the actual parking type is determined to be invalid.

[0276] The apparatus provided in this embodiment can be used to execute the technical solutions of the above method embodiments. Its implementation principle and technical effects are similar, and will not be described again here.

[0277] Figure 10 This is a schematic diagram of the parking type prediction device provided in an embodiment of this application. Figure 10 As shown, the device 100 includes an acquisition module 1001 and a processing module 1002.

[0278] The acquisition module 1001 is used to acquire the target speed curve, target relative distance curve and target parking position corresponding to the parking event when a parking event of a vehicle is detected.

[0279] The processing module 1002 is used to process the target speed curve, the target relative distance curve, and the target parking position according to the prediction model to obtain the target parking type output by the prediction model, wherein the target parking type is valid or invalid, and the prediction model is trained according to the method described in the first aspect.

[0280] In one possible design, the processing module 1002 is specifically used for:

[0281] Feature extraction is performed on the target relative distance curve to obtain target distance features;

[0282] The target distance feature is input into a distance classification model to obtain the target category output by the distance feature, wherein the target distance feature is trained according to the method described in the first aspect;

[0283] Feature extraction is performed on the target velocity curve to obtain the target velocity features;

[0284] The target speed feature, the target parking location, and the target weight corresponding to the target category are input into the prediction model to obtain the target parking type output by the prediction model.

[0285] In one possible design, the processing module 1002 is further configured to:

[0286] After the parking event ends, obtain the target parking duration corresponding to the parking event;

[0287] Based on the target parking duration and the parking duration threshold, determine the feedback parking type of the parking event;

[0288] The model parameters of the prediction model are updated based on the feedback parking type and the target parking type.

[0289] In one possible design, the processing module 1002 is further configured to:

[0290] If the target parking type is valid, then control the vehicle's internal combustion engine to shut down;

[0291] If the target parking type is invalid, the vehicle's internal combustion engine remains running.

[0292] The apparatus provided in this embodiment can be used to execute the technical solutions of the above method embodiments. Its implementation principle and technical effects are similar, and will not be described again here.

[0293] Figure 11 A schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application, such as... Figure 11 As shown, the electronic device 110 of this embodiment includes: a processor 1101 and a memory 1102; wherein

[0294] Memory 1102 is used to store computer-executed instructions;

[0295] The processor 1101 is used to execute computer execution instructions stored in the memory to implement the various steps of the prediction model training method and the parking type prediction method in the above embodiments. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0296] Alternatively, the memory 1102 can be either standalone or integrated with the processor 1101.

[0297] When the memory 1102 is set up independently, the electronic device also includes a bus 1103 for connecting the memory 1102 and the processor 1101.

[0298] This application also provides a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the prediction model training method and parking type prediction method executed by the above-mentioned electronic device.

[0299] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0300] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0301] The integrated modules implemented as software functional modules described above can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application.

[0302] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0303] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0304] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0305] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0306] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0307] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for training a prediction model, characterized in that, include: The vehicle's driving process is recorded by acquiring a first speed curve and a first relative distance curve, wherein the first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process. Based on the first speed curve, determine at least one stopping event during the driving process, and obtain the first stopping position corresponding to the stopping event; For any of the parking events, the actual parking type of the parking event is marked according to the first speed curve and the parking duration threshold. The actual parking type is either valid or invalid. When the actual parking type is valid, the internal combustion engine of the vehicle is turned off. When the actual parking type is invalid, the internal combustion engine of the vehicle remains on. The prediction model is used to process the first speed curve, the first relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model. The model parameters of the prediction model are updated based on the actual parking type and the predicted parking type.

2. The method according to claim 1, characterized in that, The step of processing the first speed curve, the first relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model includes: Based on the parking time corresponding to the parking event, a portion of the first speed curve within a preset time range before the parking time is determined as the second speed curve; A portion of the curves within a preset time range before the parking time in the first relative distance curve is determined as the second relative distance curve; The prediction model is used to process the second speed curve, the second relative distance curve, and the first parking position to obtain the predicted parking type output by the prediction model.

3. The method according to claim 2, characterized in that, The step of processing the second speed curve, the second relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model includes: The second relative distance curve is classified to determine its category; Based on the category of the second relative distance curve, determine the influence weight corresponding to the second relative distance curve; The prediction model processes the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, to obtain the predicted parking type output by the prediction model.

4. The method according to claim 3, characterized in that, The step of processing the influence weights corresponding to the second speed curve and the second relative distance curve, as well as the first parking position, according to the prediction model to obtain the predicted parking type output by the prediction model includes: Feature extraction is performed on the second velocity curve to obtain velocity features; The speed feature, the first parking position, and the influence weight are input into the prediction model to obtain the predicted parking type output by the prediction model.

5. The method according to claim 3 or 4, characterized in that, The classification process for the second relative distance curve, determining its category, includes: Feature extraction is performed on the second relative distance curve to obtain distance features; Unsupervised classification is performed based on the distance features corresponding to multiple parking events, and the multiple distance features are divided into a first category and a second category. The category of the second relative distance curve is determined based on the category to which the distance feature of the second relative distance curve belongs.

6. The method according to claim 5, characterized in that, The method further includes: A distance classification model is trained based on the distance features of the second relative distance curve and the category of the second relative distance curve. The distance classification model is used to output the category to which the distance features belong based on the distance features.

7. The method according to any one of claims 1-4, characterized in that, The step of labeling the actual parking type of the parking event based on the first speed curve and the parking duration threshold includes: Based on the first speed curve, determine the first parking duration of the parking event; If the first parking duration is greater than or equal to the parking duration threshold, then the actual parking type is determined to be valid; or, If the first parking duration is less than the parking duration threshold, then the actual parking type is determined to be invalid.

8. A parking type prediction method, characterized in that, include: When a vehicle parking event is detected, the target speed curve, target relative distance curve, and target parking position corresponding to the parking event are obtained; The target speed curve, the target relative distance curve, and the target parking position are processed according to the prediction model to obtain the target parking type output by the prediction model. The target parking type is either valid or invalid. The prediction model is trained by the method according to any one of claims 1-7.

9. The method according to claim 8, characterized in that, The process of processing the target speed curve, the target relative distance curve, and the target parking position according to the prediction model to obtain the target parking type output by the prediction model includes: Feature extraction is performed on the target relative distance curve to obtain target distance features; The target distance features are input into a distance classification model to obtain the target category output by the distance classification model, wherein the distance classification model is trained according to the method described in claim 6; Feature extraction is performed on the target velocity curve to obtain the target velocity features; The target speed feature, the target parking location, and the target influence weight corresponding to the target category are input into the prediction model to obtain the target parking type output by the prediction model.

10. The method according to claim 9, characterized in that, The method further includes: After the parking event ends, obtain the target parking duration corresponding to the parking event; Based on the target parking duration and the parking duration threshold, determine the feedback parking type of the parking event; The model parameters of the prediction model are updated based on the feedback parking type and the target parking type.

11. The method according to claim 8, characterized in that, The method further includes: If the target parking type is valid, then control the vehicle's internal combustion engine to shut down; If the target parking type is invalid, the vehicle's internal combustion engine remains running.

12. A predictive model training device, characterized in that, include: The acquisition module is used to acquire a first speed curve and a first relative distance curve corresponding to the vehicle's driving process. The first relative distance curve is used to indicate the relative distance between the vehicle and the vehicle in front during the driving process. The determining module is configured to determine at least one stopping event during the driving process based on the first speed curve; The acquisition module is also used to acquire the first parking location corresponding to the parking event; The processing module is used to, for any parking event, mark the actual parking type of the parking event according to the first speed curve and the parking duration threshold, and mark the actual parking type as valid or invalid. When the actual parking type is valid, the internal combustion engine of the vehicle is turned off, and when the actual parking type is invalid, the internal combustion engine of the vehicle remains on. The processing module is further configured to process the first speed curve, the first relative distance curve, and the first parking position according to the prediction model to obtain the predicted parking type output by the prediction model. The processing module is further configured to update the model parameters of the prediction model based on the actual parking type and the predicted parking type.

13. A parking type prediction device, characterized in that, include: The acquisition module is used to acquire the target speed curve, target relative distance curve, and target parking position corresponding to the parking event when a parking event of a vehicle is detected. The processing module is used to process the target speed curve, the target relative distance curve, and the target parking position according to the prediction model to obtain the target parking type output by the prediction model, wherein the target parking type is valid or invalid, and the prediction model is trained by the method according to any one of claims 1-7.

14. An electronic device, characterized in that, include: Memory, used to store programs; A processor for executing the program stored in the memory, wherein when the program is executed, the processor is configured to perform the method as described in any one of claims 1 to 11.

15. A computer-readable storage medium, characterized in that, Includes instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 11.

16. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 11.

Images