A vehicle control method, apparatus, device, and storage medium

Abstract

The application discloses a vehicle control method, device, equipment and storage medium, the method comprises: obtaining the current driving data of the vehicle at the current time and the historical working condition data sequence in the historical period; inputting the two to a short-time prediction model to obtain the first working condition data sequence of the vehicle in the first future period; inputting the two and the first working condition data sequence to a long-time prediction model to obtain the second working condition data sequence of the vehicle in the second future period; determining the target working condition data sequence of the vehicle in the second future period based on the first and second working condition data sequences; determining the target control instruction of the vehicle at the future time based on the target working condition data sequence and a preset target function, and controlling the vehicle according to the instruction. Through multi-time scale working condition prediction and fusion, the target control instruction is solved in combination with the preset target function, the problems of insufficient foresight and precision of single time scale prediction are overcome, and vehicle operation resource allocation is optimized.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and in particular to a vehicle control method, device, equipment and storage medium. Background Technology

[0002] Vehicle dynamic control is a core technology for ensuring driving safety and energy economy. By sensing the vehicle's operating status in real time and generating corresponding control commands, it achieves a precise match between driving intentions and vehicle responses. With the rapid development of intelligent driving technology, optimized control strategies based on predictive information have become an important direction for improving vehicle dynamic performance and energy management efficiency. The foresight and adaptability of control commands directly determine driving safety and ride comfort.

[0003] Currently, most vehicle control methods generate fixed control commands based primarily on the current state or predictions at a single time scale when making control decisions. They fail to adequately consider the dynamic characteristics of changing operating conditions at different time scales, which limits the foresight and accuracy of control commands and results in poor optimization of vehicle operating resources under complex dynamic conditions. Summary of the Invention

[0004] This application provides a vehicle control method, apparatus, device, and storage medium to address the problem of insufficient foresight and accuracy in single-time-scale predictions in the prior art.

[0005] The technical solution adopted in this application is as follows: In a first aspect, this application provides a vehicle control method, including: Obtain the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period; Input the current driving data and historical operating condition data sequence into the short-time prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period; The current driving data, historical operating condition data sequence, and first operating condition data sequence are input into the long-term prediction model to obtain the second operating condition data sequence of the vehicle in the second future time period; wherein, the second future time period is longer than the first future time period, and each future time period includes multiple time steps; Based on the first operating condition data sequence and the second operating condition data sequence, determine the target operating condition data sequence of the vehicle in the second future time period. Based on the target operating condition data sequence and the preset objective function, the target control command for the vehicle at future time is determined; wherein, the preset objective function is used to characterize the comprehensive operating resources of the vehicle in the future time period; Vehicle control is based on target control commands.

[0006] In one alternative of the first aspect, determining a target operating condition data sequence for the vehicle in a second future time period based on a first operating condition data sequence and a second operating condition data sequence includes: Extract the third operating condition data sequence of the vehicle in the first future time period from the second operating condition data sequence; The first working condition data sequence and the third working condition data sequence are weighted and fused based on preset weight coefficients to obtain a fused prediction sequence. The target operating condition data sequence is obtained based on the fused prediction sequence and the second operating condition data sequence.

[0007] In one alternative of the first aspect, based on the target operating condition data sequence and a preset objective function, the target control command for the vehicle at future times is determined, including: Based on the target operating condition data sequence and the preset target function, a target control sequence is determined; wherein, the target control sequence includes sub-control sequences corresponding to multiple time steps within the second future time period; The control command corresponding to the first sub-control sequence in the target control sequence is taken as the target control command.

[0008] In one alternative approach to the first aspect, the predefined objective function is constructed as follows: Construct at least two of the following: fuel consumption penalty, start-stop penalty, battery charge deviation penalty, and comfort penalty. Based on each penalty term and its corresponding weight coefficient, a pre-defined objective function is constructed; Set the constraints corresponding to the preset objective function.

[0009] In one alternative to the first aspect, the constraints include: The vehicle's engine power is within the preset first range; The power of the vehicle's drive motor is within the preset second range; The vehicle's battery state of charge is within the preset third range; If the vehicle's engine status changes from a stopped state to a running state, the continuous running time of the engine is greater than or equal to the first time threshold. If the vehicle's engine status changes from running to stopped, the continuous stop time of the engine is greater than or equal to the second time threshold.

[0010] In one alternative approach of the first aspect, a target control sequence is determined based on the target operating condition data sequence and a preset objective function, including: Based on the target operating condition data sequence and constraints, multiple candidate control sequences are generated; among them, the candidate control sequences include the vehicle's engine power, drive motor power, generator power, and engine operating mode at each time step in the second future time period. Based on a preset objective function, the sequence with the minimum overall operating resources is selected from multiple candidate control sequences as the target control sequence.

[0011] In one alternative of the first aspect, controlling the vehicle based on target control commands includes: Based on a preset first loss function, calculate the first loss value of the first working condition data sequence; Based on a preset second loss function, calculate the second loss value of the second working condition data sequence; If the first loss value is less than a preset first threshold and the second loss value is less than a preset second threshold, then the control operation corresponding to the target control command is executed. Otherwise, execute the control operation corresponding to the regular control command; the regular control command is determined based on the current driving data.

[0012] Secondly, this application provides a vehicle control device, comprising: The data acquisition module is used to acquire the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period; The short-term prediction module is used to input current driving data and historical operating condition data sequences into the short-term prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period. The long-term prediction module is used to input the current driving data, historical operating condition data sequence and the first operating condition data sequence into the long-term prediction model to obtain the second operating condition data sequence of the vehicle in the second future time period; wherein, the second future time period is longer than the first future time period, and each future time period includes multiple time steps. The sequence determination module is used to determine the target operating condition data sequence of the vehicle in a second future time period based on the first operating condition data sequence and the second operating condition data sequence. The instruction determination module is used to determine the target control instructions for the vehicle at future times based on the target operating condition data sequence and the preset objective function; wherein, the preset objective function is used to characterize the comprehensive operating resources of the vehicle in the future time period; The vehicle control module is used to control the vehicle based on target control commands.

[0013] In one alternative embodiment of the second aspect, the sequence determination module includes: The data extraction unit is used to extract the third operating condition data sequence of the vehicle in the first future time period from the second operating condition data sequence; The weighted fusion unit is used to perform weighted fusion of the first working condition data sequence and the third working condition data sequence based on preset weight coefficients to obtain a fused prediction sequence. The sequence integration unit is used to obtain the target operating condition data sequence based on the fused prediction sequence and the second operating condition data sequence.

[0014] In one alternative embodiment of the second aspect, the instruction determination module includes: The function solving unit is used to determine the target control sequence based on the target operating condition data sequence and the preset objective function; wherein, the target control sequence includes sub-control sequences corresponding to multiple time steps in the second future time period; The instruction determination unit is used to take the control instruction corresponding to the first sub-control sequence in the target control sequence as the target control instruction.

[0015] In one alternative approach to the second aspect, the predefined objective function is constructed as follows: Construct at least two of the following: fuel consumption penalty, start-stop penalty, battery charge deviation penalty, and comfort penalty. Based on each penalty term and its corresponding weight coefficient, a pre-defined objective function is constructed; Set the constraints corresponding to the preset objective function.

[0016] In one alternative to the second aspect, the constraints include: The vehicle's engine power is within the preset first range; The power of the vehicle's drive motor is within the preset second range; The vehicle's battery state of charge is within the preset third range; If the vehicle's engine status changes from a stopped state to a running state, the continuous running time of the engine is greater than or equal to the first time threshold. If the vehicle's engine status changes from running to stopped, the continuous stop time of the engine is greater than or equal to the second time threshold.

[0017] In one alternative embodiment of the second aspect, the instruction determination module includes: The candidate sequence generation unit is used to generate multiple candidate control sequences based on the target operating condition data sequence and constraints; wherein, the candidate control sequences include the vehicle's engine power, drive motor power, generator power and engine operating mode corresponding to each time step in the second future time period; The target sequence selection unit is used to select the sequence with the minimum comprehensive operating resources from multiple candidate control sequences as the target control sequence based on a preset objective function.

[0018] In one alternative embodiment of the second aspect, the vehicle control module includes: The first loss calculation unit is used to calculate the first loss value of the first working condition data sequence based on a preset first loss function; The second loss calculation unit is used to calculate the second loss value of the second working condition data sequence based on a preset second loss function; The control instruction execution unit is used to execute a control operation corresponding to the target control instruction if the first loss value is less than a preset first threshold and the second loss value is less than a preset second threshold. Otherwise, execute the control operation corresponding to the regular control command; the regular control command is determined based on the current driving data.

[0019] Thirdly, this application provides an electronic device including a memory and a processor. The memory is used to store computer programs or instructions that, when executed by the processor, implement the method described in the first aspect or any of the alternative solutions of the first aspect.

[0020] Fourthly, this application provides a computer-readable storage medium. The storage medium stores a computer program or instructions that, when executed by a processor, implement the method described in the first aspect or any of the alternative solutions to the first aspect.

[0021] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following: This application acquires the vehicle's current driving data and historical operating condition data sequences from historical time periods, respectively inputting them into a short-term prediction model and a long-term prediction model to obtain a first operating condition data sequence for a first future time period and a second operating condition data sequence for a second future time period. Based on the fusion of these two data sequences, a target operating condition data sequence is determined. Then, a preset objective function is used to solve for the vehicle's target control commands at future times to execute control. Unlike existing technologies that generate fixed control commands based solely on prediction results at a single time scale, this scheme achieves complementary fusion of multi-time-dimensional operating condition information through dual-scale prediction collaboration of short-term dynamics and long-term trends. This results in control command generation with superior foresight and accuracy, effectively improving the efficiency of vehicle resource optimization under complex dynamic conditions and significantly enhancing driving safety and ride comfort. Attached Figure Description

[0022] 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.

[0023] Figure 1 A flowchart of a vehicle control method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of the vehicle control device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] Embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0025] The accompanying drawings illustrate various structural schematics according to embodiments of the present disclosure. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0026] The following is a brief explanation of the terms used in this application: Vehicle communication bus: A networked data transmission architecture deployed inside a vehicle to connect various electronic control units such as powertrain, chassis, body and intelligent driving, to achieve cross-system and cross-domain vehicle-level data sharing and command interaction.

[0027] LSTM (Long Short-Term Memory) network: an improved recurrent neural network structure that selectively memorizes and updates historical information by introducing a gating mechanism. It is suitable for extracting and short-term predicting time-dependent features in vehicle historical operating condition data sequences.

[0028] Transformer encoder: A neural network encoding component based on self-attention mechanism. It captures long-distance dependencies of vehicle operating condition data by calculating the correlation weights between time steps in the sequence. It is suitable for parallel encoding and long-term prediction of multi-source time series features.

[0029] Huber loss function: A regression loss function that takes into account both mean squared error and mean absolute error. When the prediction bias is small, it manifests as mean squared error to ensure accuracy, and when the bias is large, it manifests as mean absolute error to suppress outlier interference.

[0030] With the continuous development of intelligent driving technology and the increasing demand for vehicle dynamic control, achieving precise coordination between multi-timescale condition prediction and control commands has become an important technological direction for improving driving safety and energy economy. Currently, vehicle control methods generally adopt a traditional technical architecture that generates fixed control commands based on the current state or a single timescale prediction result. In actual driving, influenced by changes in road conditions and differences in driving intentions (such as frequent acceleration and deceleration, long downhill cruising, and congested following), the vehicle's operating state exhibits both short-term drastic fluctuations and long-term trend evolution. However, when dealing with these complex driving scenarios involving multi-dimensional state perception, multi-timescale prediction, and dynamic optimization control, existing solutions still use a unified decision-making mechanism based on a single prediction mode. This limits the foresight and accuracy of control commands, results in poor optimization of vehicle operating resources under complex dynamic conditions, and fails to achieve real-time coordination between short-term dynamic response and long-term trend prediction.

[0031] To address the aforementioned technical problems, the inventive concept of this application is as follows: by acquiring the vehicle's current driving data and historical operating condition data sequences, obtaining dual-scale operating condition data sequences through short-term and long-term prediction models respectively, and determining the target operating condition data sequence based on the fusion of the two, and then combining the preset objective function to solve for the vehicle's target control commands at future times to control the vehicle, thereby achieving complementary fusion and collaborative optimization of multi-time-dimensional operating condition prediction information, overcoming the shortcomings of insufficient foresight and accuracy of control commands and improper allocation of operating resources under the single-time-scale prediction mode, ensuring real-time synchronization of multi-scale prediction and control decisions under complex dynamic operating conditions, and improving the accuracy of vehicle control commands and the efficiency of optimized allocation of operating resources.

[0032] The technical solutions of this application and how they solve the aforementioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0033] It should be noted that the executor of this application may be, but is not limited to, a vehicle controller, an intelligent driving domain controller, an on-board central computing platform, a vehicle dynamics control unit, a powertrain controller, etc.

[0034] refer to Figure 1 , Figure 1 A flowchart illustrating the vehicle control method provided in this application embodiment. Figure 1 As shown, the vehicle control method includes at least the following steps: S101: Obtain the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period.

[0035] Specifically, the vehicle's current driving data, such as speed, acceleration, battery state of charge, and pedal opening, can be collected in real time through onboard sensors and the vehicle communication bus. Historical operating condition data sequences (such as historical power data sequences and historical speed data sequences) recorded at fixed sampling periods within historical time periods can be extracted from the onboard cache to build a complete data foundation covering the current state and recent evolution trends. The historical operating condition data sequence includes multiple sets of historical operating condition data, and each set of historical operating condition data corresponds to a sampling period.

[0036] S103: Input the current driving data and historical operating condition data sequence into the short-term prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period.

[0037] Specifically, current driving data and historical operating condition data sequences can be input into a short-term prediction model (such as an LSTM network). This model learns the recent operating condition evolution patterns based on historical time series characteristics and outputs a first operating condition data sequence of the vehicle in the first future period with multiple time steps as intervals, in order to capture the short-term dynamic changes in the vehicle's operating status.

[0038] The first operating condition data sequence and the historical operating condition data sequence are data sequences of the same dimension, and also include vehicle operating status quantities such as power data and vehicle speed data, in order to maintain the consistency of the input and output feature space.

[0039] S105: Input the current driving data, historical operating condition data sequence and the first operating condition data sequence into the long-term prediction model to obtain the second operating condition data sequence of the vehicle in the second future time period.

[0040] The second future time period is longer than the first future time period, and each future time period includes multiple time steps. For example, the first future time period is the period from the current time to the next 10 seconds (e.g., 10 time steps with a step size of 1 second), and the second future time period is the period from the current time to the next 60 seconds (e.g., 60 time steps with a step size of 1 second), thus covering different prediction scales of short-term dynamics and long-term trends.

[0041] Specifically, current driving data, historical operating condition data sequences, and first operating condition data sequences can be input into a long-term prediction model (such as a Transformer encoder). The short-term prediction results are used to guide the long-term prediction, and the second operating condition data sequence of the vehicle in the second future period is output at intervals of multiple time steps, thereby obtaining operating condition evolution information over a longer period.

[0042] Similarly, the second operating condition data sequence also includes vehicle operating status quantities such as power data and vehicle speed data to support subsequent cross-time period operating condition fusion and optimization control.

[0043] S107: Based on the first operating condition data sequence and the second operating condition data sequence, determine the target operating condition data sequence of the vehicle in the second future time period.

[0044] Specifically, after obtaining the first operating condition data sequence and the second operating condition data sequence, the first operating condition data sequence and the second operating condition data sequence can be fused to form a unified and continuous target operating condition data sequence in the second future time period. This allows the accuracy of short-term prediction and the trend of long-term prediction to compensate for each other, providing a reliable basis for operating condition prediction for subsequent optimized control.

[0045] S109: Based on the target operating condition data sequence and the preset target function, determine the target control command for the vehicle at a future time.

[0046] The preset objective function is used to characterize the vehicle's comprehensive operating resources in the future time period.

[0047] Specifically, an objective function can be set to measure the comprehensive operating resources of the vehicle in the future operation process. Then, the target operating condition data sequence can be used as the expected operating state of the vehicle in the future period of time, and substituted into the preset objective function for optimization calculation, so as to obtain the vehicle target control command that optimizes the comprehensive operating resources.

[0048] S111: Control the vehicle based on target control commands.

[0049] Specifically, the target control command is sent to the corresponding actuators of the vehicle (such as the power system, braking system, and steering system), driving the vehicle to perform corresponding driving, braking, or steering actions according to the target control command, thereby achieving advance response and active control of future working conditions.

[0050] Thus, this application obtains current and historical operating condition data of the vehicle, obtains dual-scale operating condition data sequences through short-term and long-term prediction models respectively, and merges them into a target operating condition data sequence. Then, it solves the target control command based on the preset objective function to control the vehicle, realizing the complementary fusion of multi-time-scale operating condition prediction information, improving the accuracy and foresight of control command generation, and optimizing the vehicle's operating resource allocation in future time periods.

[0051] In some embodiments, the aforementioned current driving data includes vehicle status data (such as real-time vehicle speed, acceleration, battery state of charge, and power demand), driving style characteristic data (such as driving style type or quantified driving style coefficient), and environmental information data (such as road gradient, traffic density, distance to the vehicle ahead, and road type); the aforementioned historical operating condition data sequence includes historical power data and historical vehicle speed data recorded at a fixed sampling period over a past period. By fusing multi-source heterogeneous data, more comprehensive input features can be provided for subsequent dual-scale prediction models, improving prediction accuracy.

[0052] In some embodiments, the aforementioned short-term prediction model is an LSTM network, which selectively memorizes and updates the temporal dependency features in the historical operating condition data sequence through a gating mechanism to capture short-term dynamic changes in vehicle operating status; mean squared error is used as the loss function during offline training. When the system is running online, while outputting the first operating condition data sequence (corresponding to m time steps of the first future period), the m time steps closest to the current time in this historical period are used as a backtracking verification window. The actual operating condition data and their corresponding short-term prediction results for these m time steps are recorded, and the short-term prediction loss value for these m time steps is calculated to evaluate the online prediction accuracy of the short-term prediction model; wherein the number of m time steps in the backtracking verification window is equal to and corresponds one-to-one with the number of m time steps in the first future period, but they are located on the historical side and the future side of the current time, respectively. Taking historical power data as an example, the loss function is as follows:

[0053] Where L_short represents the short-term prediction loss value, m represents the number of time steps contained in the first future period, t represents the current time, Δt represents the interval of each time step, P_req_act(ti×Δt) represents the actual power demand of the most recent m steps, and P_req_s(ti×Δt) represents the predicted power demand of the most recent m steps.

[0054] In some embodiments, the long-term prediction model described above is a Transformer encoder, which calculates the correlation weights between time steps in the sequence through a self-attention mechanism to capture the long-distance dependencies of the operating condition data; the Huber function is used as the loss function during offline training. When the system is running online, while outputting the second operating condition data sequence (corresponding to n time steps of the second future period), the n time steps closest to the current time in this historical period are used as a backtracking verification window. The actual operating condition data and their corresponding long-term prediction results for these n time steps are recorded, and the long-term prediction loss value for these n time steps is calculated to evaluate the online prediction accuracy of the long-term prediction model. The number of n time steps in the backtracking verification window is equal to and corresponds one-to-one with the number of n time steps in the second future period, but they are located on the historical and future sides of the current time, respectively. Taking historical power data as an example, the loss function is as follows:

[0055] Where L_long represents the long-term prediction loss value, n represents the number of time steps contained in the second future period (n>m), t represents the current time, Δt represents the interval of each time step, P_req_act(ti×Δt) represents the actual power demand of the most recent n steps, and P_req_l(ti×Δt) represents the predicted power demand of the most recent n steps.

[0056] It should be noted that short-term power demand is affected by high-frequency factors such as driver fine-tuning and minor road surface fluctuations, and LSTM can accurately capture local evolution patterns through gating mechanisms. Long-term power demand depends on macroscopic factors (slope, traffic flow, driving style), and Transformer's self-attention mechanism can efficiently model global dependencies. Understandably, to avoid a single model weighing short-term details against long-term trends, two different models are used for prediction and then fused, ensuring both short-term accuracy and long-term trend accuracy.

[0057] In some embodiments, determining a target operating condition data sequence for the vehicle in a second future time period based on a first operating condition data sequence and a second operating condition data sequence includes: Extract the third operating condition data sequence of the vehicle in the first future time period from the second operating condition data sequence; The first working condition data sequence and the third working condition data sequence are weighted and fused based on preset weight coefficients to obtain a fused prediction sequence. The target operating condition data sequence is obtained based on the fused prediction sequence and the second operating condition data sequence.

[0058] Specifically, the future periods covered by short-term and long-term forecasts overlap, and this overlapping period can be considered a common interval where both models make predictions. Within this common interval, the prediction results of the long-term model are extracted separately and weighted together with the prediction results of the short-term model according to certain weights to obtain a fused prediction result that takes into account the advantages of both. Subsequently, this fused result can be combined with the rest of the second working condition data sequence—for example, by directly replacing the original predictions of the overlapping period or by using a smooth transition method—to obtain the target working condition data sequence that covers the entire prediction time domain.

[0059] In view of this, this embodiment performs weighted fusion of the results of dual-scale predictions within overlapping time periods, so that the accuracy of short-term predictions and the trend of long-term predictions can be mutually corrected; and various methods such as replacement, smooth connection or splicing can be flexibly selected according to the actual scenario to complete the subsequent combination, avoiding the limitations of a single combination method, and improving the flexibility and reliability of the target working condition data sequence construction.

[0060] In some embodiments, the aforementioned preset weighting coefficients can be dynamically adjusted based on the online accuracy of the dual-scale prediction. For example, when the short-term prediction loss value is below a set range, the weight of the short-term prediction result can be increased to fully utilize its local accuracy advantage; when the short-term prediction loss value is above a set range, the short-term prediction weight can be reduced accordingly, making the fusion result rely more on the trend information of the long-term prediction. As an example of fixed weights, the short-term prediction weight can be set to 0.7, the long-term prediction weight can be set to 0.3, and the sum of the two can be 1.

[0061] In some embodiments, based on a target operating condition data sequence and a preset objective function, a target control command for the vehicle at a future time is determined, including: Based on the target operating condition data sequence and the preset target function, a target control sequence is determined; wherein, the target control sequence includes sub-control sequences corresponding to multiple time steps within the second future time period; The control command corresponding to the first sub-control sequence in the target control sequence is taken as the target control command.

[0062] Specifically, the target operating condition data sequence is used as the future operating trajectory. An optimization solution is then performed using a preset objective function to obtain a series of optimal control actions corresponding to each time step within the second future time period—this is the target control sequence. Considering the continuous changes in vehicle operating conditions during operation, executing the entire sequence of control actions at once could lead to greater deviations from the actual operating conditions in later actions. Therefore, only the control command corresponding to the first time step in the target control sequence is issued as the actual target control command at each time step. When the next moment arrives, the latest data is acquired again and the solution is obtained once more, repeating this process continuously.

[0063] Therefore, this embodiment achieves rolling updates of control decisions by solving for the optimal control sequence in the complete time domain in each control cycle and executing only the first instruction. This enables the vehicle to continuously correct its control actions based on the latest operating conditions, suppresses the accumulation of prediction deviations, and improves the real-time performance and reliability of the control strategy.

[0064] In some embodiments, the preset objective function is constructed as follows: Construct at least two of the following: fuel consumption penalty, start-stop penalty, battery charge deviation penalty, and comfort penalty. Based on each penalty term and its corresponding weight coefficient, a pre-defined objective function is constructed; Set the constraints corresponding to the preset objective function.

[0065] Specifically, the preset objective function is not a single indicator, but rather transforms multiple aspects that need attention during vehicle operation into corresponding penalty terms, such as fuel consumption, frequent engine start-stop, battery charge deviation from the target value, and ride comfort. At least two of these are selected and combined. Each penalty term is assigned a corresponding weight coefficient based on its importance in the overall control objective, and a comprehensive objective function is constructed through a weighted summation. Simultaneously, constraints must be set for the solution process, such as maximum battery charging and discharging power limits and engine speed limits, to ensure that the final control commands are within the vehicle's physical capabilities.

[0066] Therefore, this embodiment transforms the multidimensional requirements of vehicle operation into quantifiable penalty terms and comprehensively weights them, avoiding the problem of sacrificing other aspects of performance by unilaterally pursuing a single performance indicator. With the setting of constraints, the optimized control commands not only meet the comprehensive optimal goal, but also satisfy the physical limitations of actual vehicle operation, thus improving the practicality and safety of the control strategy.

[0067] In some embodiments, the aforementioned fuel penalty item can be determined based on engine fuel consumption and a corresponding preset equivalent coefficient; the start-stop penalty item can be determined based on the number of engine state switching cycles and a corresponding preset equivalent coefficient; the battery charge deviation penalty item can be determined based on the deviation of the battery state of charge from the target value and a corresponding preset equivalent coefficient; and the comfort penalty item can be determined based on the change in engine power and / or drive motor power over adjacent time steps and a corresponding preset equivalent coefficient. The weighting coefficients of each of the above penalty items can be adaptively adjusted according to driving style and / or road type. For example, under an aggressive driving style, the weight of the battery maintenance item can be increased while the weight of the comfort item can be decreased; under an economical driving style, the weight of the fuel penalty item can be increased; and under urban road conditions, the weight of the start-stop item can be increased.

[0068] In some embodiments, the preset objective function is: minJ =

[0069] in: J represents the cumulative penalty score within n steps of the prediction time domain, k∈[1,n].

[0070] f_fuel(k) = w_fuel × m_fuel(k), representing the fuel penalty score; m_fuel(k) is the engine fuel consumption at time k (i.e., the kth time step), obtained by interpolating the engine universal characteristic diagram (engine power P_eng(k) at time k, engine speed n_eng(k) at time k); w_fuel is the equivalence coefficient, representing the penalty score obtained for each unit of fuel consumption, which is a preset value, such as 10.

[0071] f_stt(k)=w_start_stop×|Engine_State(k) Engine_State(k 1)| represents the score of the start-stop penalty item to prevent frequent start-stop; Engine_State(k) represents the engine state, Engine_State(k)∈{0,1}, 0 means stopped, 1 means running, Engine_State(0) is the current state; w_start_stop is the equivalent coefficient, representing the equivalent penalty score obtained for each start-stop switch, which is a preset value, such as 50.

[0072] f_soc(k) = w_soc × (SOC(k)) SOC_ref)^2 represents the SOC maintenance penalty (i.e., the battery level deviation penalty) score, keeping the battery within the ideal range; SOC(k) is the state variable, representing the estimated battery level at the k-th second; SOC_ref is the target SOC, usually the minimum balanced battery level preset by the system, such as 25 (in %); w_soc is the equivalent coefficient, representing the penalty score obtained by the square of each unit of SOC deviation, which is a preset value for calibration, such as 100.

[0073] f_comfort(k) = w_comfort_eng × (ΔP_eng(k))^2 + w_comfort_mot × (ΔP_mot(k))^2, representing the comfort penalty score to avoid power surges; P_mot(k) represents the drive motor power at time k; ΔP_eng(k) = P_eng(k) - P_eng(k-1), ΔP_mot(k) = P_mot(k) - P_mot(k-1); w_comfort_eng is an equivalent coefficient, representing the penalty score obtained by the square of the deviation of each unit of engine power, which is a preset calibration value, such as 0.5; w_comfort_mot is an equivalent coefficient, representing the penalty score obtained by the square of the deviation of each unit of motor power, which is a preset calibration value, such as 0.3.

[0074] The equivalent coefficient is usually a fixed value used to unify different physical quantities (fuel consumption, start-stop frequency, SOC deviation, power change rate) to the same cost scale, and is usually obtained through calibration tests during the development process.

[0075] ω1, ω2, ω3, and ω4 are the weight coefficients of different penalty terms, and ω1+ω2+ω3+ω4=1.

[0076] The weighting coefficients of different penalty items can be adaptively adjusted according to driving style and road type. For example, aggressive driving style focuses on power, so ω4 is smaller and ω3 is higher; economical driving style focuses on fuel economy, so ω1 is higher; highway driving focuses on range and battery maintenance, so ω1 and ω3 are higher; urban driving focuses on stop-and-go frequency, so ω2 is higher, etc.

[0077] In some embodiments, the constraints include: The vehicle's engine power is within the preset first range; The power of the vehicle's drive motor is within the preset second range; The vehicle's battery state of charge is within the preset third range; If the vehicle's engine status changes from a stopped state to a running state, the continuous running time of the engine is greater than or equal to the first time threshold. If the vehicle's engine status changes from running to stopped, the continuous stop time of the engine is greater than or equal to the second time threshold.

[0078] Specifically, these constraints define safety boundaries for control commands from multiple dimensions: engine power and drive motor power are limited to their respective reasonable preset ranges to prevent power components from operating under overload or insufficient output; battery state of charge is constrained within a preset range to avoid overcharging or over-discharging; at the same time, a minimum duration threshold is set for engine start-stop switching—if the engine has just started, it must run for at least the first time threshold (e.g., 5 seconds) before stopping, and vice versa; after the engine stops, it must maintain a stop time of at least the second time threshold (e.g., 8 seconds) to prevent the engine from repeatedly starting and stopping in a short period of time.

[0079] Therefore, this embodiment ensures that the engine and motor always operate within a safe and feasible range by setting power and state boundaries for key power components, and suppresses frequent engine switching by using start-stop time thresholds, thereby reducing mechanical wear and unnecessary energy consumption fluctuations and improving vehicle operation safety and driving comfort.

[0080] In some embodiments, determining a target control sequence based on a target operating condition data sequence and a preset objective function includes: Based on the target operating condition data sequence and constraints, multiple candidate control sequences are generated; among them, the candidate control sequences include the vehicle's engine power, drive motor power, generator power, and engine operating mode at each time step in the second future time period. Based on a preset objective function, the sequence with the minimum overall operating resources is selected from multiple candidate control sequences as the target control sequence.

[0081] Specifically, based on the target operating condition data sequence and various constraints, several feasible control schemes are first constructed as candidates. Each candidate scheme details the output power of the engine, drive motor, and generator at each future time step, as well as the engine's operating mode. Subsequently, these candidate schemes are comprehensively evaluated using a preset objective function, and the scheme that minimizes overall operating resources is selected as the final target control sequence.

[0082] Therefore, this embodiment expands the optimization range while satisfying constraints by constructing multiple candidate schemes and selecting the best one, avoiding the limitation of a single solution path falling into local suboptimal conditions, making the final control sequence better in terms of comprehensive operating resources, and improving the optimization effect of the control strategy.

[0083] In some embodiments, multiple candidate control sequences are generated based on the target operating condition data sequence and constraints. The sequence with the minimum overall operating resources is selected as the target control sequence based on a preset objective function. This can be solved using a dynamic programming algorithm. During the solution process, battery state of charge, engine state, and engine speed are used as state variables, while engine power, drive motor power, generator power, and engine operating mode commands are used as control variables. The state evolution at each time step is deduced based on the state transition equation, and the state changes are calculated based on the power balance relationship and the battery equivalent model. This allows the determination of the target control sequence that minimizes overall operating resources among the candidate control sequences.

[0084] Taking historical operating condition data sequences, including historical power data and historical vehicle speed data, as an example, the corresponding predicted power sequences and predicted vehicle speed sequences can be obtained: State variables: x(k) = [SOC(k),Engine_State(k),n_eng(k)]; Control variables: u(k) = [P_eng(k),P_motor(k),P_gen(k),Engine_Command(k)]; P_gen(k) represents the generator power at the k-th time step, and Engine_Command(k) represents the engine operating mode command at the k-th time step. State transition equation: SOC(k) = SOC(k-1) - (I_batt(k) × Δt) / Q_batt; I_batt(k) represents the battery current at the k-th time step, and Q_batt represents the battery's rated capacity. Engine_Command includes multiple operating modes such as 0 (stop), 1 (series operation), and 2 (parallel operation), which are set according to the configuration characteristics of the hybrid system. When Engine_Command(k)=0, Engine_State(k)=0, n_eng(k)=0; When Engine_Command(k)=1, Engine_State(k)=1, n_eng(k)=n_nvh(k), where n_nvh can be obtained by looking up the vehicle speed value in the predicted vehicle speed sequence. This table includes different vehicle speed values ​​and their corresponding rotational speed values. When Engine_Command(k)=2, Engine_State(k)=1, and n_eng(k) is calculated based on the predicted vehicle speed and the parallel transmission ratio; Different hybrid system configurations have certain relationships between control variables. Taking a common dual-motor (generator and drive motor) series-parallel hybrid system as an example: P_req(k) = P_mot(k) + P_eng(k) P_gen(k), where 0≤P_gen(k)≤min(P_eng(k),P_gen_max); When Engine_Command(k)=0, P_eng(k)=0; I_batt(k) is obtained by solving a first-order RC equivalent battery model, and its calculation formula is as follows:

[0085] Where V_oc and R_0 are the battery's equivalent open-circuit voltage and internal resistance, which usually vary with SOC and can be determined by looking up a table with SOC as the input. P_batt(k) can be calculated based on the state of the drive motor and the power of the generator: When P_mot(k)≥0 (motor drive), P_batt(k)=P_mot(k) / η_mot_drive-P_gen(k)×η_gen; When P_mot(k) < 0 (regenerative braking), P_batt(k) = P_mot(k) × η_mot_gen - P_gen(k) × η_gen; Among them, η_mot_drive is the drive efficiency of the drive motor, η_mot_gen is the power generation efficiency of the drive motor, and η_gen is the power generation efficiency of the generator. The above efficiencies are obtained in bench tests of the drive motor and the generator.

[0086] In some embodiments, controlling the vehicle based on target control commands includes: Based on a preset first loss function, calculate the first loss value of the first working condition data sequence; Based on a preset second loss function, calculate the second loss value of the second working condition data sequence; If the first loss value is less than a preset first threshold and the second loss value is less than a preset second threshold, then the control operation corresponding to the target control command is executed. Otherwise, execute the control operation corresponding to the regular control command; the regular control command is determined based on the current driving data.

[0087] Specifically, before issuing the target control command, a "check-up" is performed on the output quality of the two prediction models: a first loss function (such as mean squared error) is used to measure the reliability of the short-term prediction model's prediction results, yielding a first loss value; then a second loss function (such as the Huber function) is used to measure the reliability of the long-term prediction model's prediction results, yielding a second loss value. Only when both of these loss values ​​are below their respective set thresholds are the prediction results considered reliable, and the target control command calculated based on these predictions is then executed. If either prediction model fails the "check-up," it indicates that the current prediction may be unreliable. In this case, predictive control is not used recklessly; instead, the system reverts to the conventional control command directly determined based on the current real-time driving data as a safety precaution.

[0088] Based on this, this embodiment verifies the reliability of the dual-scale prediction results, enables optimized control to improve operating efficiency when the prediction is reliable, and automatically switches to conventional control to ensure safety when the prediction is inaccurate. This achieves adaptive selection between predictive control and conventional control, avoids the risk of inappropriate control commands due to model prediction deviations, and improves the safety and robustness of vehicle control.

[0089] Based on the same technical concept, this application also provides a vehicle control device, see reference. Figure 2 , Figure 2 This is a schematic diagram of the vehicle control device provided in an embodiment of this application. Figure 2 As shown, the vehicle control device includes at least a data acquisition module 201, a short-term prediction module 202, a long-term prediction module 203, a sequence determination module 204, an instruction determination module 205, and a vehicle control module 206, wherein: The data acquisition module 201 is used to acquire the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period; The short-term prediction module 202 is used to input the current driving data and historical operating condition data sequence into the short-term prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period. The long-term prediction module 203 is used to input the current driving data, historical operating condition data sequence and the first operating condition data sequence into the long-term prediction model to obtain the second operating condition data sequence of the vehicle in the second future time period; wherein, the second future time period is longer than the first future time period, and each future time period includes multiple time steps. The sequence determination module 204 is used to determine the target operating condition data sequence of the vehicle in a second future time period based on the first operating condition data sequence and the second operating condition data sequence. The instruction determination module 205 is used to determine the target control instructions for the vehicle at future times based on the target operating condition data sequence and the preset objective function; wherein, the preset objective function is used to characterize the comprehensive operating resources of the vehicle in the future time period; The vehicle control module 206 is used to control the vehicle based on target control commands.

[0090] In some embodiments, the sequence determination module 204 includes: The data extraction unit is used to extract the third operating condition data sequence of the vehicle in the first future time period from the second operating condition data sequence; The weighted fusion unit is used to perform weighted fusion of the first working condition data sequence and the third working condition data sequence based on preset weight coefficients to obtain a fused prediction sequence. The sequence integration unit is used to obtain the target operating condition data sequence based on the fused prediction sequence and the second operating condition data sequence.

[0091] In some embodiments, the instruction determining module 205 includes: The function solving unit is used to determine the target control sequence based on the target operating condition data sequence and the preset objective function; wherein, the target control sequence includes sub-control sequences corresponding to multiple time steps in the second future time period; The instruction determination unit is used to take the control instruction corresponding to the first sub-control sequence in the target control sequence as the target control instruction.

[0092] In some embodiments, the preset objective function is constructed as follows: Construct at least two of the following: fuel consumption penalty, start-stop penalty, battery charge deviation penalty, and comfort penalty. Based on each penalty term and its corresponding weight coefficient, a pre-defined objective function is constructed; Set the constraints corresponding to the preset objective function.

[0093] In some embodiments, the constraints include: The vehicle's engine power is within the preset first range; The power of the vehicle's drive motor is within the preset second range; The vehicle's battery state of charge is within the preset third range; If the vehicle's engine status changes from a stopped state to a running state, the continuous running time of the engine is greater than or equal to the first time threshold. If the vehicle's engine status changes from running to stopped, the continuous stop time of the engine is greater than or equal to the second time threshold.

[0094] In some embodiments, the instruction determining module 205 includes: The candidate sequence generation unit is used to generate multiple candidate control sequences based on the target operating condition data sequence and constraints; wherein, the candidate control sequences include the vehicle's engine power, drive motor power, generator power and engine operating mode corresponding to each time step in the second future time period; The target sequence selection unit is used to select the sequence with the minimum comprehensive operating resources from multiple candidate control sequences as the target control sequence based on a preset objective function.

[0095] In some embodiments, the vehicle control module 206 includes: The first loss calculation unit is used to calculate the first loss value of the first working condition data sequence based on a preset first loss function; The second loss calculation unit is used to calculate the second loss value of the second working condition data sequence based on a preset second loss function; The control instruction execution unit is used to execute a control operation corresponding to the target control instruction if the first loss value is less than a preset first threshold and the second loss value is less than a preset second threshold. Otherwise, execute the control operation corresponding to the regular control command; the regular control command is determined based on the current driving data.

[0096] It should be noted that this vehicle control device can be used to implement any of the above-described method embodiments.

[0097] Based on the same technical concept, this application also provides an electronic device, see reference. Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device includes a memory 301 and a processor 302. The memory 301 is used to store computer instructions; when the processor 302 executes the computer instructions, it implements any of the above-described method embodiments.

[0098] The memory 301 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium may be an internal storage unit of an electronic device, such as the hard disk or memory of the electronic device. In other embodiments, the computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, secure digital card (SD card), flash memory card, etc., equipped on the electronic device. Of course, the computer-readable storage medium may include both internal storage units and external storage devices of the electronic device. In this embodiment, the computer-readable storage medium is typically used to store the operating system and various application software installed on the electronic device, such as the program code of the vehicle control method in the embodiment. In addition, the computer-readable storage medium may also be used to temporarily store various types of data that have been output or will be output.

[0099] In some embodiments, processor 302 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other chip. Processor 302 is typically used to control the overall operation of the electronic device, such as performing control and processing related to data interaction or communication with other entities. In this embodiment, processor 302 is used to run program code stored in memory 301 or process data.

[0100] Based on the same technical concept, this application also provides a computer-readable storage medium, which includes a computer program or instructions stored in the storage medium. When the computer program or instructions are executed by a processing device, they implement any of the above-described method embodiments. Further details can be found in the method embodiments, which will not be repeated here. In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium can be an internal storage unit of an electronic device, such as the hard disk or memory of the electronic device. In other embodiments, the computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, a secure digital card (SD card), a flash memory card, etc., equipped on the electronic device. Of course, the computer-readable storage medium can also include both internal storage units and external storage devices of the electronic device. In this embodiment, the computer-readable storage medium is typically used to store the operating system and various application software installed on the electronic device, such as the program code of the vehicle control method in the embodiment. Furthermore, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or will be output.

[0101] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0102] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A vehicle control method, characterized in that, include: Obtain the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period; The current driving data and the historical operating condition data sequence are input into the short-time prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period. The current driving data, the historical operating condition data sequence, and the first operating condition data sequence are input into a long-term prediction model to obtain the second operating condition data sequence of the vehicle in a second future time period; wherein the second future time period is longer than the first future time period, and each future time period includes multiple time steps; Based on the first operating condition data sequence and the second operating condition data sequence, determine the target operating condition data sequence of the vehicle in the second future time period; Based on the target operating condition data sequence and the preset objective function, the target control command for the vehicle at a future time is determined; wherein, the preset objective function is used to characterize the comprehensive operating resources of the vehicle in the future time period; The vehicle is controlled based on the target control command.

2. The method according to claim 1, characterized in that, The step of determining the target operating condition data sequence of the vehicle in the second future time period based on the first operating condition data sequence and the second operating condition data sequence includes: Extract the third operating condition data sequence of the vehicle in the first future time period from the second operating condition data sequence; The first working condition data sequence and the third working condition data sequence are weighted and fused based on preset weighting coefficients to obtain a fused prediction sequence. Based on the fused prediction sequence and the second operating condition data sequence, the target operating condition data sequence is obtained.

3. The method according to claim 1, characterized in that, The step of determining the target control command for the vehicle at a future time based on the target operating condition data sequence and a preset target function includes: Based on the target operating condition data sequence and the preset target function, a target control sequence is determined; wherein, the target control sequence includes sub-control sequences corresponding to multiple time steps within the second future time period; The control instruction corresponding to the first sub-control sequence in the target control sequence is taken as the target control instruction.

4. The method according to claim 3, characterized in that, The preset objective function is constructed in the following manner: Construct at least two of the following: fuel consumption penalty, start-stop penalty, battery charge deviation penalty, and comfort penalty. Based on each penalty term and its corresponding weight coefficient, the preset objective function is constructed; Set the constraints corresponding to the preset objective function.

5. The method according to claim 4, characterized in that, The constraints include: The engine power of the vehicle is within a preset first range; The power of the vehicle's drive motor is within a preset second range; The battery state of charge of the vehicle is within a preset third range; If the engine status of the vehicle switches from a stopped state to a running state, the continuous running time of the engine is greater than or equal to a first time threshold. If the engine state of the vehicle switches from the running state to the stopped state, the continuous stop time of the engine is greater than or equal to the second time threshold.

6. The method according to claim 4, characterized in that, The step of determining the target control sequence based on the target operating condition data sequence and the preset target function includes: Based on the target operating condition data sequence and the constraints, multiple candidate control sequences are generated; wherein, the candidate control sequences include the vehicle's engine power, drive motor power, generator power, and engine operating mode corresponding to each time step in the second future time period; Based on the preset objective function, the sequence with the minimum comprehensive operating resources is selected from the multiple candidate control sequences as the target control sequence.

7. The method according to claim 1, characterized in that, Controlling the vehicle based on the target control command includes: Based on a preset first loss function, calculate the first loss value of the first working condition data sequence; Based on a preset second loss function, calculate the second loss value of the second working condition data sequence; If the first loss value is less than a preset first threshold and the second loss value is less than a preset second threshold, then the control operation corresponding to the target control command is executed. Otherwise, execute the control operation corresponding to the regular control command; wherein the regular control command is determined based on the current driving data.

8. A vehicle control device, characterized in that, include: The data acquisition module is used to acquire the vehicle's current driving data at the current moment and the historical operating condition data sequence within the historical time period; The short-term prediction module is used to input the current driving data and the historical operating condition data sequence into the short-term prediction model to obtain the first operating condition data sequence of the vehicle in the first future time period. The long-term prediction module is used to input the current driving data, the historical operating condition data sequence and the first operating condition data sequence into the long-term prediction model to obtain the second operating condition data sequence of the vehicle in a second future time period; wherein, the second future time period is longer than the first future time period, and each future time period includes multiple time steps. The sequence determination module is used to determine the target operating condition data sequence of the vehicle in the second future time period based on the first operating condition data sequence and the second operating condition data sequence. The instruction determination module is used to determine the target control instruction of the vehicle at a future time based on the target operating condition data sequence and a preset objective function; wherein, the preset objective function is used to characterize the comprehensive operating resources of the vehicle in the future time period; A vehicle control module is used to control the vehicle based on the target control command.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory being used to store computer programs or instructions; when the computer programs or instructions are executed by the processor, the method of any one of claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, which, when executed by a processor, implement the method of any one of claims 1-7.

Images