Whole vehicle quality determination method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a vehicle quality determination method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: periodically acquiring multiple groups of target characteristic values of a vehicle longitudinal dynamic model within a preset time period; updating a plurality of model parameters of the preset number of initial deep neural network models for three times based on the plurality of groups of target feature values; and determining the whole vehicle mass of the target vehicle based on the deep neural network model updated for the third time. According to the method provided by the invention, the occupation of on-chip resources can be reduced, and the convergence duration in the training process of the deep neural network model can be effectively shortened.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining vehicle weight. Background Technology

[0002] Commercial vehicles require torque adjustment based on road conditions, load, and driving demands during operation to achieve optimal power output and fuel efficiency. High-performance vehicle weight estimation algorithms provide accurate weight parameters, serving as characteristic value references for torque distribution optimization schemes and ensuring a balance between power and fuel economy under different operating conditions. Therefore, efficient and accurate vehicle weight estimation algorithms are of great significance to commercial vehicles.

[0003] Currently, most manufacturers' optimization algorithms use real-time optimization methods; however, these algorithms consume a significant amount of on-chip resources. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium, and computer program product for determining vehicle mass that can reduce the occupation of on-chip resources, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for determining the weight of a vehicle, the method comprising:

[0006] Multiple sets of target feature values ​​of the vehicle's longitudinal dynamics model are periodically acquired within a preset time period. The target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope.

[0007] The multiple sets of target feature values ​​are input into a preset number of initial deep neural network models to update the multiple model parameters in the initial deep neural network models for the first time.

[0008] The multiple sets of target feature values ​​are input into each of the first updated deep neural network models to obtain the first model output quality. Based on the first model output quality, the actual vehicle weight, the preset cross probability, and the preset cross weight, the multiple model parameters in the first updated deep neural network model are updated a second time.

[0009] Based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round, a third update is performed on multiple model parameters in the deep neural network model after the second update.

[0010] The overall vehicle mass of the target vehicle is determined based on the third updated deep neural network model.

[0011] In one embodiment, the second update of multiple model parameters in the first updated deep neural network model, based on the first model output quality, the actual vehicle quality, the preset crossover probability, and the preset crossover weight, includes:

[0012] Based on the output quality of the first model and the actual quality of the vehicle, a corresponding first fitness value is obtained;

[0013] Based on all first fitness values, obtain the selection probability of each deep neural network model after the first update;

[0014] Based on the selection probability, the preset crossover probability, and the preset crossover weight, a second update is performed on multiple model parameters in the deep neural network model after the first update.

[0015] In one embodiment, the second update of multiple model parameters in the first updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight includes:

[0016] Based on the selection probability, a preset number of deep neural network models are selected from the first updated deep neural network models according to a preset number of selections, with one deep neural network model selected each time.

[0017] Based on the selection order of the deep neural network models, two adjacent selected deep neural network models are grouped together;

[0018] Based on the grouping results, the preset crossover probability, and the preset crossover weight, a second update is performed on multiple model parameters in the deep neural network model after the first update.

[0019] In one embodiment, the second update of multiple model parameters in the first updated deep neural network model based on the grouping results, the preset crossover probability, and the preset crossover weight includes:

[0020] Based on the preset crossover probability, for any deep neural network model in each group of deep neural network models, the model parameters in any deep neural network model are updated based on the model parameters in the any deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight value.

[0021] In one embodiment, determining the overall vehicle mass based on the third updated deep neural network model includes:

[0022] If the current training round is not less than the preset training round, the multiple sets of target feature values ​​are input into the deep neural network model after each third update to obtain the second model output quality, and the corresponding second fitness value is obtained based on the second model output quality and the actual vehicle quality.

[0023] The deep neural network model updated after the third time, corresponding to the minimum second fitness value, is determined as the target deep neural network model;

[0024] The overall vehicle mass is determined based on the target deep neural network model.

[0025] In one embodiment, the method further includes:

[0026] If the current training round is less than the preset training round, the deep neural network model after the third update is determined as the new initial deep neural network model. The process returns to the step of inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models and continues until the current training round is not less than the preset training round.

[0027] Secondly, this application also provides a vehicle weight determination device, the device comprising:

[0028] The acquisition module is used to periodically acquire multiple sets of target feature values ​​of the vehicle's longitudinal dynamics model within a preset time period. The target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope.

[0029] The first update module is used to input the multiple sets of target feature values ​​into a preset number of initial deep neural network models to update the multiple model parameters in the initial deep neural network models for the first time.

[0030] The second update module is used to input the multiple sets of target feature values ​​into each deep neural network model after the first update to obtain the first model output quality, and based on the first model output quality, the actual vehicle quality, the preset cross probability and the preset cross weight, to perform a second update on multiple model parameters in the deep neural network model after the first update.

[0031] The third update module is used to update multiple model parameters in the deep neural network model after the second update for a third time based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round.

[0032] The determination module is used to determine the overall vehicle mass of the target vehicle based on the third updated deep neural network model.

[0033] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method in any of the above embodiments.

[0034] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the methods in any of the above embodiments.

[0035] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the methods in any of the above embodiments.

[0036] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining vehicle mass periodically acquire multiple sets of target feature values ​​from the vehicle's longitudinal dynamics model within a preset time period. These target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope. The multiple sets of target feature values ​​are input into a preset number of initial deep neural network models to update multiple model parameters in the initial deep neural network models for the first time. The multiple sets of target feature values ​​are then input into each of the first-updated deep neural network models to obtain the first model output mass. Based on the first model output mass, the actual vehicle mass, preset crossover probabilities, and preset crossover weights, multiple model parameters in the first-updated deep neural network models are updated a second time. Based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round, multiple model parameters in the second-updated deep neural network models are updated a third time. The vehicle mass of the target vehicle is determined based on the third-updated deep neural network model. The method provided in this application can reduce the occupation of on-chip resources and effectively shorten the convergence time during the training process of deep neural network models. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 is a flowchart illustrating a method for determining vehicle weight in one embodiment;

[0039] Figure 2 is a schematic diagram of a deep neural network model in one embodiment;

[0040] Figure 3 is a flowchart illustrating the second update method in one embodiment;

[0041] Figure 4 is a flowchart of a method for determining vehicle weight in another embodiment;

[0042] Figure 5 is a structural block diagram of a vehicle weight determination device in one embodiment;

[0043] Figure 6 is an internal structure diagram of a computer device in one embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0045] In one embodiment, as shown in Figure 1, a method for determining vehicle weight is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0046] S102. Periodically acquire multiple sets of target feature values ​​of the vehicle's longitudinal dynamics model within a preset time period. The target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope.

[0047] In the case of longitudinal driving, the longitudinal dynamics model of the vehicle is shown in the following equation:

[0048]

[0049] In the formula, For rolling resistance, For air resistance, For slope resistance, To represent acceleration resistance, G is the total weight of the vehicle. The rolling resistance coefficient is... For slope, Here, A is the air resistance coefficient, A is the frontal area, and v is the current vehicle speed. is the conversion factor for vehicle rotational mass, where m is the total vehicle mass.

[0050] In the above formula, the frontal area and various coefficients are determined by the vehicle's physical structure and can be constants. Apart from these, the only characteristic values ​​of the vehicle's longitudinal dynamics model are the vehicle mass, speed, acceleration, gradient, and driving force. Since the purpose of this embodiment is to estimate the vehicle mass, the driving wheel speed, driven wheel speed, driving force, and road gradient are determined as the target characteristic values ​​input to the deep neural network model. In this embodiment, L sets of target characteristic values ​​are periodically acquired within a preset time period. These L sets of data represent the continuous process of the vehicle's operation over a certain period.

[0051] The driven wheel speed can characterize the vehicle's travel speed. The change in driven wheel speed in the L sets of data can characterize acceleration. Furthermore, when the driving wheel and driven wheel speeds differ, it can also characterize the vehicle's current degree of slippage, indirectly representing the rolling resistance coefficient. In this way, the characteristics of neural networks can be used to approximate a nonlinear model of vehicle dynamics, with the output being the vehicle's mass.

[0052] S104. Input multiple sets of target feature values ​​into a preset number of initial deep neural network models to perform the first update of multiple model parameters in the initial deep neural network models.

[0053] The deep neural network model in this embodiment is shown in Figure 2. The deep neural network model in the figure includes an input layer, multiple hidden layers and an output layer. By inputting L sets of target feature values ​​into the trained deep neural network model, the corresponding vehicle mass can be obtained.

[0054] Multiple model parameters include multiple weight values ​​and multiple bias values. The model parameters are extracted into a vector form as shown in the following formula:

[0055]

[0056] In the formula, and These represent the weights and biases involved in the deep neural network model. The size of this vector is determined by the number of neural network nodes and the number of hidden layers. The number of neural network nodes is 4L, but since the data in the L groups are only different in terms of collection time, their physical meanings are somewhat correlated, so the number of hidden layers does not need to be too large.

[0057] The process of updating the model parameters for the first time is the backpropagation optimization process of the initial neural network model. During this process, the weight and bias values ​​in the model have changed. The number of iterations of backpropagation optimization can be represented by iter. The larger the value of iter, the more accurate the backpropagation optimization process is and the greater the computing power required.

[0058] S106. Input multiple sets of target feature values ​​into each of the first updated deep neural network models to obtain the first model output quality, and based on the first model output quality, the actual vehicle quality, the preset cross probability and the preset cross weight, perform a second update on multiple model parameters in the first updated deep neural network model.

[0059] The process for determining the output quality of the first model after each initial deep neural network model update is shown in the following formula:

[0060]

[0061] In the formula, The output quality of the first model is given by input, which is the target feature value. (DNN) n For a deep neural network model, where, N is the preset quantity.

[0062] S108. Based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round, perform a third update on multiple model parameters in the deep neural network model after the second update.

[0063] The mutation probability refers to the probability of updating each model parameter in the deep neural network model after each second update, and the upper limit of the mutation probability refers to the maximum value that the mutation probability can be obtained. The model parameters can be updated based on the mutation probability or the upper limit of the mutation probability, or the model parameters can be updated based on both the mutation probability and the upper limit of the mutation probability at the same time. This application does not specifically limit this. As the number of training rounds increases, both the mutation probability and the upper limit of the mutation probability will decrease.

[0064] For example, the current mutation probability is used This indicates that the current upper limit of the mutation probability is represented by... It means that, among them, , Based on If it is determined that a certain model parameter in a model needs to be updated for the third time, then this model parameter is compared with... The product between them is determined as the updated value of the model parameters.

[0065] S110. Determine the overall vehicle mass of the target vehicle based on the third updated deep neural network model.

[0066] First, the target deep neural network model is determined based on the third updated deep neural network model. Then, the corresponding target feature values ​​of the target vehicle are input into the target deep neural network model to obtain the overall vehicle mass of the target vehicle.

[0067] In the aforementioned method for determining the overall vehicle mass, multiple sets of target feature values ​​from the vehicle's longitudinal dynamics model are periodically acquired within a preset time period. These target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope. The multiple sets of target feature values ​​are then input into a preset number of initial deep neural network models to update multiple model parameters in the initial deep neural network models for the first time. The multiple sets of target feature values ​​are then input into each of the first-updated deep neural network models to obtain the first model output mass. Based on the first model output mass, the actual vehicle mass, preset crossover probabilities, and preset crossover weights, multiple model parameters in the first-updated deep neural network models are updated a second time. Based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round, multiple model parameters in the second-updated deep neural network models are updated a third time. Finally, the overall vehicle mass of the target vehicle is determined based on the third-updated deep neural network model. The method provided in this application can reduce the occupation of on-chip resources and effectively shorten the convergence time during the training process of deep neural network models.

[0068] In some embodiments, as shown in Figure 3, based on the first model output quality, the actual vehicle quality, the preset crossover probability, and the preset crossover weight, a second update is performed on multiple model parameters in the deep neural network model after the first update, including:

[0069] S302. Obtain the corresponding first fitness value based on the output quality of the first model and the actual quality of the vehicle.

[0070] S304. Based on all first fitness values, obtain the selection probability of each deep neural network model after the first update.

[0071] S306. Based on the selection probability, preset crossover probability, and preset crossover weight, perform a second update on multiple model parameters in the deep neural network model after the first update.

[0072] The fitness value is used to characterize the performance of a deep neural network model. A smaller fitness value indicates that the model's output value is closer to the actual value, and therefore, the better the model's performance. The process of determining the first fitness value is shown in the following formula:

[0073]

[0074] In the formula, Cost is the first fitness value, and output is the actual mass of the vehicle.

[0075] The first fitness value of the nth deep neural network model can be represented by Cost. nThe expression is as follows: First, the sum of the reciprocals of all first fitness values ​​is calculated, and then the selection probability of each deep neural network model after the first update is determined based on the sum of the reciprocals.

[0076]

[0077]

[0078] In the formula, It is the sum of the reciprocals. To select the probability.

[0079] In this embodiment, by determining the first fitness value and selection probability of each deep neural network model after the first update, the subsequent model selection process becomes more accurate.

[0080] In some embodiments, based on the selection probability, preset crossover probability, and preset crossover weights, a second update is performed on multiple model parameters in the deep neural network model after the first update. This includes: selecting a preset number of deep neural network models from the deep neural network model after the first update according to the selection probability and a preset number of selections, selecting one deep neural network model at a time; grouping adjacent selected deep neural network models into a group based on the selection order of the deep neural network models; and updating multiple model parameters in the deep neural network model after the first update based on the grouping results, preset crossover probability, and preset crossover weights.

[0081] The selection process uses a roulette wheel principle to filter the deep neural network models after the first update. The preset number of selections and the preset number of models are used. In each selection, one model is chosen from all the deep neural network models after the first update. The same model can be selected in different selection attempts, with models having a higher probability of being chosen being selected more frequently. After selecting the preset number of deep neural network models, they are sorted according to the selection order. Starting with the first model in the sorted results, adjacent selected deep neural network models are grouped together. For example, if the sorted results include models A, B, C, and D, then models A and B are grouped together, and models C and D are grouped together.

[0082] In this embodiment, by filtering and grouping the deep neural network models after the first update, the results of the subsequent second update are made more accurate.

[0083] In some embodiments, based on the grouping results, preset crossover probabilities, and preset crossover weights, a second update is performed on multiple model parameters in the deep neural network model after the first update. This includes: based on the preset crossover probabilities, for any deep neural network model in each group of deep neural network models, updating the model parameters in any deep neural network model based on the model parameters in any deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight values.

[0084] The preset crossover probability refers to the probability that the parameters of two models at corresponding positions within the same group will be updated based on preset crossover weights. The higher the preset crossover probability, the greater the probability that the parameters of the two models at corresponding positions will be updated. All model parameters within the same group have the same preset crossover probability and preset crossover weights.

[0085] Specifically, the preset crossover probability is used This indicates that the preset cross-weight values ​​are used. It means that, among them, , Based on If it is determined that two model parameters at corresponding positions in two models within the same group need to be updated, for any one of the model parameters, calculate the relationship between that model parameter and the given parameter. The first product between them, and calculate the other model parameters with The second product between the first and second products determines the updated value of any model parameter.

[0086] In this embodiment, the deep neural network model is updated a second time based on the preset crossover probability and preset crossover weight value, which can further optimize the model parameters.

[0087] In some embodiments, determining the overall vehicle mass of the target vehicle based on the third-updated deep neural network model includes: inputting multiple sets of target feature values ​​into each third-updated deep neural network model, where the current training round is not less than a preset training round, to obtain the second model output mass, and obtaining the corresponding second fitness value based on the second model output mass and the actual vehicle mass; determining the third-updated deep neural network model corresponding to the smallest second fitness value as the target deep neural network model; and determining the overall vehicle mass based on the target deep neural network model.

[0088] If the current training round is not less than the preset training round, it means that the maximum number of iterations has been reached. At this time, multiple sets of target values ​​are input into the deep neural network model after the third update, and the deep neural network model after the third update is backpropagated again to obtain the second fitness value. Finally, the target deep neural network model is determined based on the second fitness value.

[0089] In this embodiment, when the current training round is not less than the preset training round, the target deep neural network model is determined based on the second fitness value, so that the determined target deep neural network model has better performance.

[0090] In some embodiments, the method further includes: if the current training round is less than a preset training round, determining the third updated deep neural network model as the new initial deep neural network model, returning to the step of inputting multiple sets of target feature values ​​into a preset number of initial deep neural network models and continuing to execute until the current training round is not less than the preset training round.

[0091] If the current training round is less than the preset training round, it means that the maximum number of iterations has not been reached, and the model training is not sufficient. Therefore, the deep neural network model needs to be further optimized.

[0092] In this embodiment, if the current training round has not yet reached the maximum number of iterations, the model can continue to be optimized to ensure that the performance of the final target deep neural network model meets the requirements.

[0093] In one embodiment, another method for determining vehicle instructions is provided. The flowchart of this method is shown in Figure 4. The method includes the following steps: First, offline model training is performed by building a neural network. Then, the bias and weight coefficients of the neural network are optimized using the characteristics of a genetic algorithm to avoid local optima caused by random point distribution. Finally, the trained neural network is run on an actual vehicle. This approach avoids real-time optimization on a real vehicle while maintaining low complexity and fast convergence speed, demonstrating good performance and application potential.

[0094] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0095] Based on the same inventive concept, this application also provides a vehicle weight determination apparatus for implementing the above-described vehicle weight determination method. The solution provided by this apparatus is similar to the implementation described in the above-described method; therefore, the specific limitations in one or more embodiments of the vehicle weight determination apparatus provided below can be found in the limitations of the vehicle weight determination method described above, and will not be repeated here.

[0096] In an exemplary embodiment, as shown in FIG5, a vehicle weight determination device 500 is provided, including: an acquisition module 501, a first update module 502, a second update module 503, a third update module 504, and a determination module 505, wherein:

[0097] The acquisition module 501 is used to periodically acquire multiple sets of target feature values ​​of the vehicle's longitudinal dynamics model within a preset time period. The target feature values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope.

[0098] The first update module 502 is used to input the multiple sets of target feature values ​​into a preset number of initial deep neural network models to perform the first update of multiple model parameters in the initial deep neural network models.

[0099] The second update module 503 is used to input the multiple sets of target feature values ​​into each of the first updated deep neural network models to obtain the first model output quality, and based on the first model output quality, the actual vehicle quality, the preset cross probability and the preset cross weight, to perform a second update on multiple model parameters in the first updated deep neural network model.

[0100] The third update module 504 is used to update multiple model parameters in the deep neural network model after the second update for the third time based on the current mutation probability and the upper limit of the current mutation probability corresponding to the current training round.

[0101] The determination module 505 is used to determine the overall vehicle mass of the target vehicle based on the third updated deep neural network model.

[0102] In some embodiments, the second update module 503 includes:

[0103] The first acquisition submodule is used to acquire a corresponding first fitness value based on the output quality of the first model and the actual quality of the vehicle.

[0104] The second acquisition submodule is used to acquire the selection probability of each deep neural network model after the first update based on all first fitness values.

[0105] The update submodule is used to perform a second update on multiple model parameters in the deep neural network model after the first update, based on the selection probability, the preset crossover probability, and the preset crossover weight.

[0106] In some embodiments, updating a submodule includes:

[0107] The selection unit is used to select a preset number of deep neural network models from the first updated deep neural network models based on the selection probability and according to a preset number of selections, with one deep neural network model selected each time.

[0108] The grouping unit is used to group two adjacent selected deep neural network models into a group based on the selection order of the deep neural network models.

[0109] The update unit is used to update multiple model parameters in the deep neural network model after the first update a second time based on the grouping results, the preset crossover probability, and the preset crossover weight.

[0110] In some embodiments, the updating unit is further configured to update the model parameters of any deep neural network in each group of deep neural network models based on the preset cross-probability, the model parameters in the any deep neural network model, the corresponding model parameters in another deep neural network model, and the preset cross-weight value.

[0111] In some embodiments, the determining module 505 is further configured to, when the current training round is not less than a preset training round, input the multiple sets of target feature values ​​into each third-updated deep neural network model to obtain the second model output quality, and obtain the corresponding second fitness value based on the second model output quality and the actual vehicle quality; determine the third-updated deep neural network model corresponding to the smallest second fitness value as the target deep neural network model; and determine the overall vehicle quality based on the target deep neural network model.

[0112] In some embodiments, the determining module 505 is further configured to, when the current training round is less than the preset training round, determine the deep neural network model after the third update as the new initial deep neural network model, return to the step of inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models and continue to execute until the current training round is not less than the preset training round.

[0113] Each module in the aforementioned vehicle weight determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0114] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram is shown in Figure 6. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for determining vehicle weight.

[0115] Those skilled in the art will understand that the structure shown in Figure 6 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0116] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: periodically acquiring multiple sets of target feature values ​​of a vehicle longitudinal dynamics model within a preset time period, the target feature values ​​including the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope; inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models to perform a first update on multiple model parameters in the initial deep neural network models; inputting the multiple sets of target feature values ​​into each of the first-updated deep neural network models to obtain a first model output quality, and performing a second update on multiple model parameters in the first-updated deep neural network models based on the first model output quality, the actual vehicle mass, a preset crossover probability, and a preset crossover weight; performing a third update on multiple model parameters in the second-updated deep neural network models based on the current mutation probability and the current mutation probability upper limit corresponding to the current training round; and determining the overall vehicle mass of the target vehicle based on the third-updated deep neural network model.

[0117] In one embodiment, when the processor executes a computer program, it performs a second update on multiple model parameters in the first-updated deep neural network model based on the first model output quality, the actual vehicle quality, a preset crossover probability, and a preset crossover weight. This includes: obtaining corresponding first fitness values ​​based on the first model output quality and the actual vehicle quality; obtaining the selection probability corresponding to each first-updated deep neural network model based on all first fitness values; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight.

[0118] In one embodiment, the processor, when executing a computer program, performs a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight. This includes: selecting a preset number of deep neural network models from the first-updated deep neural network model according to the selection probability and a preset number of selections, selecting one deep neural network model at a time; grouping adjacent selected deep neural network models into a group based on the selection order; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the grouping result, the preset crossover probability, and the preset crossover weight.

[0119] In one embodiment, when the processor executes a computer program, it performs a second update on multiple model parameters in the first updated deep neural network model based on the grouping result, the preset crossover probability, and the preset crossover weight. This includes: based on the preset crossover probability, for any deep neural network model in each group of deep neural network models, updating the model parameters in any deep neural network model based on the model parameters in that deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight value.

[0120] In one embodiment, the determination of the overall vehicle mass of a target vehicle based on a third-updated deep neural network model implemented by the processor when executing a computer program includes: when the current training round is not less than a preset training round, inputting the multiple sets of target feature values ​​into each third-updated deep neural network model to obtain a second model output mass, and obtaining a corresponding second fitness value based on the second model output mass and the actual vehicle mass; determining the third-updated deep neural network model corresponding to the smallest second fitness value as the target deep neural network model; and determining the overall vehicle mass based on the target deep neural network model.

[0121] In one embodiment, the method implemented by the processor when executing the computer program further includes: if the current training round is less than the preset training round, determining the third updated deep neural network model as the new initial deep neural network model, returning to the step of inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models and continuing to execute until the current training round is not less than the preset training round.

[0122] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program performs the following steps: periodically acquiring multiple sets of target feature values ​​of a vehicle longitudinal dynamics model within a preset time period, the target feature values ​​including the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope; inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models to perform a first update on multiple model parameters in the initial deep neural network models; inputting the multiple sets of target feature values ​​into each of the first-updated deep neural network models to obtain a first model output quality, and performing a second update on multiple model parameters in the first-updated deep neural network models based on the first model output quality, the actual vehicle mass, a preset crossover probability, and a preset crossover weight; performing a third update on multiple model parameters in the second-updated deep neural network models based on the current mutation probability and the current mutation probability upper limit corresponding to the current training round; and determining the overall vehicle mass of the target vehicle based on the third-updated deep neural network model.

[0123] In one embodiment, when the computer program is executed by the processor, it performs a second update on multiple model parameters in the first-updated deep neural network model based on the first model output quality, the actual vehicle quality, a preset crossover probability, and a preset crossover weight. This includes: obtaining corresponding first fitness values ​​based on the first model output quality and the actual vehicle quality; obtaining a selection probability for each first-updated deep neural network model based on all first fitness values; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight.

[0124] In one embodiment, when a computer program is executed by a processor, it performs a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight. This includes: selecting a preset number of deep neural network models from the first-updated deep neural network model according to the selection probability and a preset number of selections, selecting one deep neural network model at a time; grouping adjacent selected deep neural network models into a group based on the selection order; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the grouping result, the preset crossover probability, and the preset crossover weight.

[0125] In one embodiment, when a computer program is executed by a processor, it performs a second update on multiple model parameters in the first updated deep neural network model based on the grouping results, the preset crossover probability, and the preset crossover weight. This includes: based on the preset crossover probability, for any deep neural network model in each group of deep neural network models, updating the model parameters in any deep neural network model based on the model parameters in that deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight value.

[0126] In one embodiment, the determination of the overall vehicle mass of a target vehicle based on a third-updated deep neural network model, implemented by a computer program executed by a processor, includes: when the current training round is not less than a preset training round, inputting the multiple sets of target feature values ​​into each third-updated deep neural network model to obtain a second model output mass, and obtaining a corresponding second fitness value based on the second model output mass and the actual vehicle mass; determining the third-updated deep neural network model corresponding to the smallest second fitness value as the target deep neural network model; and determining the overall vehicle mass based on the target deep neural network model.

[0127] In one embodiment, the method implemented when the computer program is executed by the processor further includes: if the current training round is less than the preset training round, determining the third updated deep neural network model as the new initial deep neural network model, returning to the step of inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models and continuing to execute until the current training round is not less than the preset training round.

[0128] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps: periodically acquiring multiple sets of target feature values ​​of a vehicle's longitudinal dynamics model within a preset time period, the target feature values ​​including the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope; inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models to perform a first update on multiple model parameters in the initial deep neural network models; inputting the multiple sets of target feature values ​​into each of the first-updated deep neural network models to obtain a first model output quality, and performing a second update on multiple model parameters in the first-updated deep neural network models based on the first model output quality, the actual vehicle mass, a preset crossover probability, and a preset crossover weight; performing a third update on multiple model parameters in the second-updated deep neural network models based on the current mutation probability and the current mutation probability upper limit corresponding to the current training round; and determining the overall vehicle mass of the target vehicle based on the third-updated deep neural network model.

[0129] In one embodiment, when the computer program is executed by the processor, it performs a second update on multiple model parameters in the first-updated deep neural network model based on the first model output quality, the actual vehicle quality, a preset crossover probability, and a preset crossover weight. This includes: obtaining corresponding first fitness values ​​based on the first model output quality and the actual vehicle quality; obtaining a selection probability for each first-updated deep neural network model based on all first fitness values; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight.

[0130] In one embodiment, when a computer program is executed by a processor, it performs a second update on multiple model parameters in the first-updated deep neural network model based on the selection probability, the preset crossover probability, and the preset crossover weight. This includes: selecting a preset number of deep neural network models from the first-updated deep neural network model according to the selection probability and a preset number of selections, selecting one deep neural network model at a time; grouping adjacent selected deep neural network models into a group based on the selection order; and performing a second update on multiple model parameters in the first-updated deep neural network model based on the grouping result, the preset crossover probability, and the preset crossover weight.

[0131] In one embodiment, when a computer program is executed by a processor, it performs a second update on multiple model parameters in the first updated deep neural network model based on the grouping results, the preset crossover probability, and the preset crossover weight. This includes: based on the preset crossover probability, for any deep neural network model in each group of deep neural network models, updating the model parameters in any deep neural network model based on the model parameters in that deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight value.

[0132] In one embodiment, the determination of the overall vehicle mass of a target vehicle based on a third-updated deep neural network model, implemented by a computer program executed by a processor, includes: when the current training round is not less than a preset training round, inputting the multiple sets of target feature values ​​into each third-updated deep neural network model to obtain a second model output mass, and obtaining a corresponding second fitness value based on the second model output mass and the actual vehicle mass; determining the third-updated deep neural network model corresponding to the smallest second fitness value as the target deep neural network model; and determining the overall vehicle mass based on the target deep neural network model.

[0133] In one embodiment, the method implemented when the computer program is executed by the processor further includes: if the current training round is less than the preset training round, determining the third updated deep neural network model as the new initial deep neural network model, returning to the step of inputting the multiple sets of target feature values ​​into a preset number of initial deep neural network models and continuing to execute until the current training round is not less than the preset training round.

[0134] 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, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0136] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0137] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining vehicle mass, characterized in that: The method comprises: Periodically acquiring multiple groups of target characteristic values ​​of the vehicle longitudinal dynamics model within a preset time period, wherein the target characteristic values ​​include the vehicle's driving wheel speed, driven wheel speed, driving force, and road slope; Inputting the multiple groups of target feature values ​​into a preset number of initial deep neural network models to perform a first update on multiple model parameters in the initial deep neural network models; Inputting the multiple groups of target feature values ​​into each deep neural network model after the first update to obtain a first model output quality, and performing a second update on multiple model parameters in the deep neural network model after the first update based on the first model output quality, the actual vehicle quality, a preset crossover probability, and a preset crossover weight; Based on the current mutation probability and the current mutation probability upper limit corresponding to the current training round, performing a third update on the multiple model parameters in the deep neural network model after the second update; Determine the vehicle mass of the target vehicle based on the deep neural network model after the third update.

2. The method according to claim 1, characterized in that The second updating of the plurality of model parameters in the deep neural network model after the first update based on the first model output mass, the actual mass of the vehicle, the preset crossover probability and the preset crossover weight comprises: Acquire a corresponding first fitness value based on the first model output mass and the actual mass of the vehicle; Obtaining the corresponding selection probability of each first updated deep neural network model based on all first fitness values; Based on the selection probability, the preset crossover probability and the preset crossover weight, a plurality of model parameters in the deep neural network model after the first update are updated for a second time.

3. The method according to claim 2, characterized in that The second updating of the plurality of model parameters in the deep neural network model after the first update based on the selection probability, the preset crossover probability and the preset crossover weight comprises: Based on the selection probability, selecting a preset number of deep neural network models from the deep neural network model after the first update according to a preset number of selections, and selecting one deep neural network model each time; Based on the selection order of the deep neural network models, two adjacent selected deep neural network models are grouped together; Based on the grouping result, the preset crossover probability and the preset crossover weight, a plurality of model parameters in the deep neural network model after the first update are updated for a second time.

4. The method according to claim 3, characterized in that: The second updating of the plurality of model parameters in the deep neural network model after the first update based on the grouping result, the preset crossover probability and the preset crossover weight comprises: Based on the preset crossover probability, for any deep neural network model in each group of deep neural network models, the model parameters in any deep neural network model are updated based on the model parameters in the any deep neural network model, the corresponding model parameters in another deep neural network model, and the preset crossover weight value.

5. The method according to claim 1, characterized in that The determining the vehicle mass of the target vehicle based on the deep neural network model after the third update includes: When the current training round is not less than the preset training round, the multiple groups of target feature values ​​are input into each third updated deep neural network model to obtain a second model output quality, and a corresponding second fitness value is obtained based on the second model output quality and the actual mass of the vehicle; Determine the third updated deep neural network model corresponding to the minimum second fitness value as the target deep neural network model; The vehicle mass is determined based on the target deep neural network model.

6. The method according to claim 5, characterized in that The method further comprises: When the current training round is less than the preset training round, the deep neural network model after the third update is determined as the new initial deep neural network model, and the process returns to the step of inputting the multiple groups of target feature values ​​into a preset number of initial deep neural network models and continues until the current training round is not less than the preset training round.

7. A vehicle mass determination device, characterized in that: The device comprises: An acquisition module, used to periodically acquire multiple groups of target characteristic values ​​of the vehicle longitudinal dynamics model within a preset time period, wherein the target characteristic values ​​include the driving wheel speed, the driven wheel speed, the driving force and the road slope of the vehicle; A first updating module, configured to input the multiple groups of target feature values ​​into a preset number of initial deep neural network models to perform a first update on multiple model parameters in the initial deep neural network models; A second updating module is used to input the multiple groups of target feature values ​​into each deep neural network model after the first update to obtain a first model output quality, and based on the first model output quality, the actual vehicle quality, a preset crossover probability and a preset crossover weight, perform a second update on multiple model parameters in the deep neural network model after the first update; A third updating module, configured to perform a third update on a plurality of model parameters in the deep neural network model after the second update based on a current mutation probability and a current mutation probability upper limit corresponding to a current training round; A determination module is used to determine the vehicle mass of the target vehicle based on the deep neural network model after the third update.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images