A method and apparatus for optimizing vehicle dynamics data under multiple driving scenarios

Abstract

This invention relates to the field of vehicle power output optimization technology, and discloses a method and apparatus for optimizing vehicle power data under multiple driving scenarios. The invention predicts optimized values ​​for power output, torque demand, and energy recovery time using a target neural network, and generates a loss function between the optimized value of each of these parameters and its corresponding actual value. Then, using global and local optimization algorithms, the optimized values ​​for power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario. Therefore, this invention significantly improves the accuracy of power data prediction in multiple driving scenarios by using global and local optimization algorithms to further optimize the prediction results output by the target neural network, while also ensuring that the drive system can perform actions optimally in dynamic environments.

Description

Technical Field

[0001] This invention relates to the field of vehicle power output optimization technology, specifically to a method and apparatus for optimizing vehicle power data under multiple driving scenarios. Background Technology

[0002] In intelligent transportation and the automotive industry, drive system optimization technology plays a significant role in improving vehicle energy efficiency, power response, and overall driving performance. Furthermore, with the development of intelligent driving technology, the performance of vehicle drive systems not only affects vehicle power output and fuel / battery efficiency but also directly relates to driving safety and the driving experience. Therefore, effectively optimizing vehicle drive systems has become an important research direction.

[0003] Currently, traditional vehicle drive system optimization methods typically use preset rules or fixed parameters, but these methods struggle to adapt to dynamic and complex driving scenarios. Vehicles need to handle various driving conditions during operation, such as urban congestion, highway driving, and hill climbing, making it impossible for a single drive system optimization strategy to address diverse needs. To improve the adaptability of drive systems, modern vehicles are usually equipped with multiple sensors to collect real-time data on vehicle operating status, road conditions, and environmental conditions. This data provides a data-driven foundation for drive system optimization to a certain extent.

[0004] In related technologies, the power output optimization method of the drive system is generally optimized according to a fixed mode. However, the parameter setting method of the fixed mode has obvious limitations when dealing with multiple driving scenarios. The optimization results are not accurate enough, and thus cannot provide sufficient dynamic adjustment capabilities, resulting in poor performance of vehicle power response and energy efficiency optimization. Summary of the Invention

[0005] In view of this, the present invention provides a method and apparatus for optimizing the power data of a vehicle in multiple driving scenarios, in order to solve the problem that optimizing the power output of the drive system according to fixed parameters has limitations and the optimization results are not accurate enough, which leads to the impact on the vehicle's power response and energy efficiency.

[0006] According to a first aspect, the present invention provides a method for optimizing power output data of a vehicle under multiple driving scenarios, the method comprising:

[0007] In the current driving scenario, the predicted power output based on vehicle road condition data, the predicted torque demand based on traffic data and driving environment data, and the predicted energy recovery time based on historical energy recovery time data are obtained respectively.

[0008] Obtain the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario;

[0009] Based on the actual power output power, obtain the first execution parameter corresponding to the predicted power output power; based on the actual torque demand, obtain the second execution parameter corresponding to the predicted torque demand; based on the actual torque demand, obtain the third execution parameter corresponding to the predicted energy recovery time.

[0010] The system predicts the power output power using a target neural network, generates an optimized power output power value using a first execution parameter, predicts the torque demand using a target neural network, generates an optimized torque demand value using a second execution parameter, predicts the energy recovery time using a target neural network, generates an optimized energy recovery time value using a third execution parameter, and generates a loss function between the optimized value of each of the above items and the actual value corresponding to that optimized value. The loss function is used to update the weight parameter values ​​of the target neural network.

[0011] The power output, torque demand, and energy recovery time are further optimized within the driving cycle under the current driving scenario using both global and local optimization algorithms.

[0012] In some optional implementations, based on the actual power output power value, the first execution parameter corresponding to the predicted power output power value is obtained and executed using the following formula:

[0013] ΔP(t)=λ P ·(P actual (t)-P pred (t))

[0014] P exec (t)=P pred (t)+ΔP(t)

[0015] Where ΔP(t) is the first deviation between the actual power output and the predicted power output, and λ P P is the first weighting parameter. actual (t) represents the actual value of the power output, P pred (t) represents the predicted power output, P exec (t) is the first execution parameter corresponding to the predicted power output value.

[0016] In some optional implementations, a second execution parameter corresponding to the predicted torque demand value is obtained based on the actual torque demand value, and executed using the following formula:

[0017] ΔB(t)=λ B ·(B factual (t)-B pred (t))

[0018] B exec (t)=Bpred (t)+ΔB(t)

[0019] Where ΔB(t) is the second deviation between the actual torque demand and the predicted torque demand, and λ B B is the second weighting parameter. actual (t) represents the actual torque demand, B pred (t) represents the predicted torque demand, B pred (t) represents the predicted torque demand, B exec (t) is the second execution parameter.

[0020] In some optional implementations, the predicted energy recovery time based on historical energy recovery time data is calculated using the following formula:

[0021] ΔE(t)=λ E ·(E actual (t)-E pred (t))

[0022] E exec (t)=E pred (t)+ΔE(t)

[0023] Where ΔE(t) is the third deviation between the actual value of the energy recovery time and the predicted value of the energy recovery time, λ E E is the third weighting parameter. actual (t) represents the actual energy recovery time, E pred (t) represents the predicted energy recovery time, E exec (t) is the third execution parameter.

[0024] In some optional implementations, a loss function is generated between the optimized value of each of the above items and the corresponding actual value. This loss function is used to update the weight parameter values ​​of the target neural network, and is executed using the following formula:

[0025]

[0026] in, For loss function, For each of the above predicted values, y i The actual value corresponding to each of the above optimization values, Ω(f) k ) represents the regularization term. For the error loss term, f k The target neural network uses the Kth tree of a decision tree, where m is the number of samples used to train the target neural network, M is the total number of target neural networks, and w h (t) represents the weighting parameter for historical vehicle detection data, w r (t) represents the weighting parameter of the current vehicle detection data. This is a predicted value based on historical data. The predicted value is based on the current data. This represents the optimized value for each of the above items.

[0027] In some optional implementations, the power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a global optimization algorithm, executed according to the following formula:

[0028]

[0029] Where, r global,1 (t) is the reward function of the global optimization algorithm, A1 is the set of actions to be executed by the global optimization algorithm, and π global,1 (t) represents the optimal result corresponding to the optimal action taken by the global optimization algorithm.

[0030] In some optional implementations, the power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a local optimization algorithm, executed according to the following formula:

[0031]

[0032] r local,2 (t)=-|M actual (t)-M pred (t)|

[0033] Where, r local,2 (t) is the reward function of the local optimization algorithm, A2 is the set of actions to be performed by the local optimization algorithm, and π local,2 (t) represents the optimal result corresponding to the optimal action taken by the local optimization algorithm, M. actual (t) represents the actual value of each of the above terms, M pred (t) represents the predicted value for each of the above items.

[0034] In some optional implementations, the vehicle power data optimization method in multiple driving scenarios in this disclosure embodiment further includes: combining the optimization results of local optimization algorithms and global optimization algorithms through the following formula;

[0035]

[0036] Where, π i (t) represents the weights of the optimization results of the collaborative global optimization algorithm and the local optimization algorithm in the current driving scenario, r scene,i (t) represents the reward value of the optimization results of the collaborative global optimization algorithm and the local optimization algorithm in the current driving scenario, r global,3(t) represents the final result after collaboration, and N represents the number of collaborating objects.

[0037] Secondly, the present invention provides a power output data optimization device for a vehicle under multiple driving scenarios, the device comprising:

[0038] The first acquisition module is used to acquire, in the current driving scenario, the predicted power output value based on vehicle road condition data, the predicted torque demand value based on traffic data and driving environment data, and the predicted energy recovery time value based on historical energy recovery time data.

[0039] The second acquisition module is used to acquire the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario.

[0040] The third acquisition module is used to acquire the first execution parameter corresponding to the predicted value of power output power based on the actual value of power output power, acquire the second execution parameter corresponding to the predicted value of torque demand based on the actual value of torque demand, and acquire the third execution parameter corresponding to the predicted value of energy recovery time based on the actual value of torque demand.

[0041] The data prediction module is used to predict the power output power through the target neural network, generate the optimized power output power through the first execution parameter, predict the torque demand through the target neural network, generate the optimized torque demand through the second execution parameter, predict the energy recovery time through the target neural network, generate the optimized energy recovery time through the third execution parameter, and generate a loss function between the optimized value of each of the above items and the actual value corresponding to the optimized value. The loss function is used to update the weight parameter values ​​of the target neural network.

[0042] The data optimization module is used to collaboratively optimize the power output, torque demand, and energy recovery time values ​​within the driving cycle under the current driving scenario using both global and local optimization algorithms.

[0043] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the vehicle power output data optimization method in multiple driving scenarios as described in the first aspect or any corresponding embodiment.

[0044] The technical solution of this invention has the following advantages:

[0045] This invention relates to the field of vehicle power output optimization technology, and discloses a method and apparatus for optimizing vehicle power data under multiple driving scenarios. The invention uses a target neural network to predict power output power, generates optimized power output power values ​​using a first execution parameter, predicts torque demand using a target neural network, generates optimized torque demand values ​​using a second execution parameter, predicts energy recovery time using a target neural network, and generates optimized energy recovery time values ​​using a third execution parameter. It also generates a loss function between the optimized value and the corresponding actual value for each of these parameters, using the loss function to update the weight parameter values ​​of the target neural network. Furthermore, it uses global and local optimization algorithms to further optimize the power output power, torque demand, and energy recovery time values ​​within the driving cycle of the current driving scenario. Therefore, this invention utilizes a target neural network combined with real-time power output-related data, predicted data, and the execution results corresponding to the predicted data to obtain optimized data. Then, it uses global and local optimization algorithms to further optimize the predicted results output by the target neural network, significantly improving the accuracy of vehicle power data prediction under multiple driving scenarios. Simultaneously, it ensures that the drive system can execute actions optimally in dynamic environments. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating a method for optimizing vehicle power output data in multiple driving scenarios according to an embodiment of the present invention.

[0048] Figure 2 This is a simplified structural block diagram of a multilayer network structure according to an embodiment of the present invention;

[0049] Figure 3 This is a flowchart illustrating another method for optimizing power output data of a vehicle in multiple driving scenarios according to an embodiment of the present invention.

[0050] Figure 4 This is a flowchart illustrating another method for optimizing power output data of a vehicle in multiple driving scenarios according to an embodiment of the present invention.

[0051] Figure 5 This is a flowchart illustrating a method for optimizing power output data of a vehicle in multiple driving scenarios according to an embodiment of the present invention.

[0052] Figure 6 This is a simplified block diagram illustrating how a vehicle collects real-time vehicle detection data in multiple driving scenarios according to an embodiment of the present invention.

[0053] Figure 7 This is another method for optimizing vehicle power output data under multiple driving scenarios according to an embodiment of the present invention.

[0054] Figure 8 This is a structural block diagram of a vehicle power output data optimization device under multiple driving scenarios according to an embodiment of the present invention;

[0055] Figure 9 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] According to an embodiment of the present invention, a method for optimizing power output data of a vehicle in multiple driving scenarios is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0058] This embodiment provides a method for optimizing vehicle power output data under multiple driving scenarios, which can be used in computer devices such as mobile phones, tablets, desktop computers, laptops, servers, etc. Figure 1 This is a flowchart of a method for optimizing vehicle power output data under multiple driving scenarios according to an embodiment of the present invention. The process includes the following steps:

[0059] Step S101: Under the current driving scenario, obtain the predicted power output value based on vehicle road condition data, the predicted torque demand value based on traffic data and driving environment data, and the predicted energy recovery time value based on historical energy recovery time data.

[0060] Specifically, in the current driving scenario (urban congestion, highway driving, or hill start), the predicted power output based on vehicle road condition data is predicted through the first sub-network; the predicted torque demand based on traffic and driving environment data is predicted through the second sub-network; and the predicted energy recovery time based on historical energy recovery time data is predicted through the third sub-network. The first, second, and third sub-networks constitute a multi-layered network structure. For example... Figure 2 The diagram shown is a simplified structural block diagram of the multi-layer network structure in an embodiment of this disclosure. Figure 2 The network consists of a first subnetwork, a second subnetwork, and a third subnetwork, which together form a multi-layered network structure. The first subnetwork serves as the bottom layer and is used to process basic data. The second subnetwork serves as the middle layer and is used to integrate road condition data, environmental data, and traffic data. The third subnetwork serves as the top layer and is used to process historical data.

[0061] Step S102: Obtain the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario.

[0062] Specifically, various sensors installed on the vehicle monitor in real time the actual power output, torque demand, and energy recovery time of the vehicle under the current driving scenario. For example, a power output sensor monitors the actual power output P of the vehicle during actual operation. actual (t), and compared with the predicted power output value P. pred (t) is compared, and the comparison result is used to identify the difference between the actual power output and the predicted power output, thereby determining whether the power output needs to be adjusted. For example, the actual torque demand B of the vehicle in actual operation is monitored in real time by a torque demand sensor. factual (t), and compared with the torque demand forecast B. pred (t) is compared, and the comparison result is used to identify the difference between the actual torque demand and the predicted torque demand, thereby determining whether torque adjustment is needed. For example, the actual value of the energy recovery time E of the vehicle in actual operation is monitored in real time by an energy recovery time sensor. actual (t), and compared with the predicted energy recovery time E. pred The comparison (t) is used to identify the difference between the actual and predicted energy recovery times, thereby determining whether the energy recovery time needs to be adjusted.

[0063] Step S103: Based on the actual power output power value, obtain the first execution parameter corresponding to the predicted power output power value; based on the actual torque demand value, obtain the second execution parameter corresponding to the predicted torque demand value; based on the actual torque demand value, obtain the third execution parameter corresponding to the predicted energy recovery time value.

[0064] In some optional implementations, based on the actual power output power value, the first execution parameter corresponding to the predicted power output power value is obtained and executed using the following formula:

[0065] ΔP(t)=λ P ·(P actual (t)-P pred (t))

[0066] P exec (t)=P pred (t)+ΔP(t)

[0067] Where ΔP(t) is the first deviation between the actual power output and the predicted power output, and λ P P is the first weighting parameter. actual (t) represents the actual value of the power output, P pred (t) represents the predicted power output, P exec (t) is the first execution parameter corresponding to the predicted power output value.

[0068] In some optional implementations, a second execution parameter corresponding to the predicted torque demand value is obtained based on the actual torque demand value, and executed using the following formula:

[0069] ΔB(t)=λ B ·(B factual (t)-B pred (t))

[0070] B exec (t)=B pred (t)+ΔB(t)

[0071] Where ΔB(t) is the second deviation between the actual torque demand and the predicted torque demand, and λ B B is the second weighting parameter. actual (t) represents the actual torque demand, B pred (t) represents the predicted torque demand, B pred (t) represents the predicted torque demand, B exec (t) is the second execution parameter.

[0072] In some optional implementations, the predicted energy recovery time based on historical energy recovery time data is calculated using the following formula:

[0073] ΔE(t)=λ E ·(E actual (t)-E pred (t))

[0074] E exec (t)=E pred (t)+ΔE(t)

[0075] Where ΔE(t) is the third deviation between the actual value of the energy recovery time and the predicted value of the energy recovery time, λ E E is the third weighting parameter. actual (t) represents the actual energy recovery time, E pred (t) represents the predicted energy recovery time, E exec (t) is the third execution parameter.

[0076] Step S104: The target neural network predicts the power output power prediction value and generates the power output power optimization value using the first execution parameter; the target neural network predicts the torque demand prediction value and generates the torque demand optimization value using the second execution parameter; the target neural network predicts the energy recovery time prediction value and generates the energy recovery time optimization value using the third execution parameter; and generates a loss function between the optimization value of each of the above items and the actual value corresponding to the optimization value. The loss function is used to update the weight parameter values ​​of the target neural network.

[0077] Specifically, the predicted power output value and its corresponding first execution parameter output by the first sub-network, the predicted torque demand value and its corresponding second execution parameter output by the second sub-network, and the predicted energy recovery time value and its corresponding third feature parameter output by the third sub-network are used as input features to the target neural network to predict the optimized power output value, optimized torque demand value, and optimized energy recovery time value. This target neural network can be an XGBoost neural network. This target neural network is trained using historical samples. Traditional methods adjust vehicle power output based solely on the predicted value, while this embodiment combines the predicted value with the corresponding execution parameters, creating a bidirectional closed-loop feedback structure in the target neural network. The model is further optimized based on the execution results, and the drive system is continuously optimized during operation by continuously adjusting the model parameters. For example, when the drive system detects that the predicted power output value is insufficient, the drive system increases the power output in real time and feeds this execution result back to the target neural network model, allowing it to make corresponding corrections in the next prediction.

[0078] In some optional implementations, a loss function is generated between the optimized value of each of the above items and the corresponding actual value. This loss function is used to update the weight parameter values ​​of the target neural network, and is executed using the following formula:

[0079]

[0080] in, For loss function, For each of the above predicted values, y i The actual value corresponding to each of the above optimization values, Ω(f) k ) represents the regularization term. For the error loss term, f k The target neural network uses the Kth tree of a decision tree, where m is the number of samples used to train the target neural network, M is the total number of target neural networks, and w h (t) represents the weighting parameter for historical vehicle detection data, w r (t) represents the weighting parameter of the current vehicle detection data. This is a predicted value based on historical data. The predicted value is based on the current data. This represents the optimized value for each of the above items.

[0081] Specifically, the weighting parameters of historical vehicle detection data and current vehicle detection data mentioned above are adaptively adjusted according to environmental changes. For example, when a vehicle is driving under stable driving conditions, the weighting parameter w of the historical vehicle detection data... h (t) will increase, indicating reliance on past trend predictions; while when the environment or road conditions change suddenly, the weight of the current vehicle detection data w will increase. r (t) will increase to ensure the model can quickly respond to new data changes. The weight parameter w of historical vehicle detection data... h (t) and the current vehicle detection data weight parameter w r (t) can be adjusted in real time using an adaptive optimization algorithm, which can be either gradient descent or Bayesian optimization. The adaptive optimization algorithm adjusts the weight parameters based on the gradient of the prediction error in the loss function, ensuring that the target prediction neural network achieves optimal prediction performance and response capability under different driving scenarios.

[0082] In some optional implementations, during the process of the target neural network predicting the predicted value of the power output power and the first execution parameter generating the optimized value of the power output power, the prediction result of the current prediction is passed to the next prediction through the following cyclic feedback formula;

[0083] P next (t+1)=P pred (t+1)+η1·(P exec(t)-P pred (t))

[0084] Among them, P next (t+1) represents the next optimized power output value, P exec (t) represents the first execution parameter corresponding to the predicted power output value, P exec (t) represents the first execution parameter corresponding to the predicted power output value, P exec (t)-P pred (t) represents the first deviation between the actual power output power value and the predicted power output power value, and η1 is the influence weight of the current power output power optimization prediction on the next power output power optimization prediction.

[0085] In some optional implementations, during the process of the target neural network predicting the torque demand forecast value and the second execution parameter generating the torque demand optimization value, the current prediction result is passed to the next prediction through the following cyclic feedback formula;

[0086] B next (t+1)=B pred (t+1)+η2·(B exec (t)-B pred (t))

[0087] Among them, B next (t+1) represents the optimized torque demand value for the next iteration, B exec (t) represents the second execution parameter corresponding to the predicted torque demand value, B exec (t) represents the second execution parameter corresponding to the predicted torque demand value, B exec (t)-B pred (t) represents the second deviation between the actual value of the current torque demand and the predicted value of the current torque demand, and η2 is the weight of the influence of the current torque demand prediction on the next torque demand prediction.

[0088] In some optional implementations, during the process of the target neural network predicting the predicted value of the energy recovery time and the third execution parameter generating the optimized value of the energy recovery time, the prediction result of the current prediction is passed to the next prediction through the following cyclic feedback formula;

[0089] E next (t+1)=E pred (t+1)+η3·(E exec (t)-E pred (t))

[0090] Among them, E next (t+1) is the optimized value for the next energy recovery time, E exec(t) represents the third execution parameter corresponding to the predicted energy recovery time, E exec (t) represents the third execution parameter corresponding to the predicted energy recovery time, E exec (t)-E pred (t) represents the third deviation between the actual value of the current energy recovery time and the predicted value of the current energy recovery time, and η3 is the weight of the influence of the current energy recovery time prediction on the next energy recovery time prediction.

[0091] Step S105: Through global optimization algorithm and local optimization algorithm, the power output power optimization value, torque demand optimization value and energy recovery time optimization value are jointly optimized within the driving cycle under the current driving scenario.

[0092] In some optional implementations, the power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a global optimization algorithm, executed according to the following formula:

[0093]

[0094] Where, r global,1 (t) is the reward function of the global optimization algorithm, A1 is the set of actions to be executed by the global optimization algorithm, and π global,1 (t) represents the optimal result corresponding to the optimal action taken by the global optimization algorithm.

[0095] Specifically, a global optimization algorithm is used to formulate a global task plan and further coordinate power output, torque demand, and energy recovery time in the current driving scenario (urban congestion scenario, highway driving scenario, and hill driving scenario).

[0096] For example, the drive system sets a global objective by combining vehicle road condition data, traffic data, driving environment data, and energy recovery time data. During the global optimization process, the drive system employs a reinforcement learning algorithm to evaluate the reward or benefit E[global(t)] brought by taking different global actions in the current state, and selects the action A1 that maximizes the reward. global (t).

[0097] In this embodiment, a global optimization algorithm is used to calculate a global reward function for target planning. During the target planning process, the drive system dynamically selects the optimal global action based on the current driving scenario. For example, in urban congestion scenarios, priority is given to energy recovery from smooth driving and frequent braking, while in high-speed driving scenarios, the focus is on the energy efficiency of continuous power output. The reinforcement learning algorithm described above continuously adjusts the execution actions in the global action set using historical and real-time data, ensuring that the drive system maintains the best balance between power and energy efficiency in different driving scenarios. Therefore, the global optimization algorithm provides global scheduling and planning for the entire driving process, ensuring that the vehicle can take optimal actions in various complex driving environments, achieving efficient energy management and power distribution.

[0098] In some optional implementations, the power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a local optimization algorithm, as expressed by the following formula:

[0099]

[0100] r local,2 (t)=-|M actual (t)-M pred (t)|

[0101] Where, r local,2 (t) is the reward function of the local optimization algorithm, A2 is the set of actions to be performed by the local optimization algorithm, and π local,2 (t) represents the optimal result corresponding to the optimal action taken by the local optimization algorithm, M. actual (t) represents the actual value of each of the above terms, M pred (t) represents the predicted value for each of the above items.

[0102] Specifically, the local optimization algorithm formulates the target and plans the local task, and further coordinates the power output, torque demand and energy recovery time under the current driving scenario (urban congestion scenario, highway driving scenario, slope driving scenario).

[0103] In the above formula, the local reward function π local,2 (t) is used to evaluate the performance of each local task. This local reward function is based on the actual output M. actual (t) and predicted value M pred The deviation of (t) is further calculated, which reflects the adjustment effect of the local task. The local reward function π is set. local,2The goal of (t) is to minimize the deviation between the actual output and the predicted value, enabling the drive system to execute local tasks more accurately in future adjustments. At each time point in vehicle operation, the local optimization algorithm performs local optimization based on the feedback from the local reward function, adjusting local actions to improve the performance of immediate tasks. For example, if the current power output deviation is large, the drive system will adjust the power output accordingly based on this feedback to ensure the accuracy of subsequent tasks. This optimization process iterates repeatedly, continuously updating the adjustment results, enabling the drive system to execute actions in the optimal way in a dynamic environment, thereby achieving optimal power output, torque demand, and energy recovery.

[0104] In some optional implementations, the optimization results of local optimization algorithms and global optimization algorithms are combined using the following formula;

[0105]

[0106] Where, π i (t) represents the weights of the optimization results of the collaborative global optimization algorithm and the local optimization algorithm in the current driving scenario, r scene,i (t) represents the reward value of the optimization results of the collaborative global optimization algorithm and the local optimization algorithm in the current driving scenario, r global,3 (t) represents the final result after collaboration, and N represents the number of collaborating objects.

[0107] Specifically, the reward function r global,3 (t) is used to coordinate the optimization results of global and local optimizations during the driving cycle, so that the drive system can maintain optimal energy efficiency and power output in different driving scenarios (urban congestion scenario, highway driving scenario, and hill driving scenario).

[0108] This embodiment of the disclosure ensures that the execution of local target tasks is consistent with the execution of global target tasks by adjusting the optimization results of local optimization algorithms. By coordinating global and local optimization algorithms, the optimal power output, torque demand, and energy recovery time are ultimately obtained.

[0109] This embodiment provides a method for optimizing vehicle power output data under multiple driving scenarios. It can be used with computer devices such as mobile phones, tablets, desktop computers, laptops, and servers. Step S101 involves obtaining, in the current driving scenario, a predicted power output value based on vehicle road condition data, a predicted torque demand value based on traffic data and driving environment data, and a predicted energy recovery time value based on historical energy recovery time data. Figure 3 As shown, it includes:

[0110] Step S301: Obtain the current detection data, historical average power demand data, and historical energy recovery time data of the vehicle in the current driving scenario. The current detection data includes: vehicle operation data, vehicle road condition data, traffic data, and driving environment data.

[0111] Specifically, vehicle operating data includes: acceleration, speed, and engine speed, where acceleration... v(t) is the vehicle speed, t is time, and a(t) is the acceleration, which is detected by an accelerometer, while the speed is detected by a vehicle speed sensor. Vehicle road condition data includes: gradient, which can be detected by cameras, LiDAR sensors, and millimeter-wave radar sensors. Traffic data includes: traffic flow. Where, N vehicles (t) represents the current road area per unit area A. lane The number of vehicles on the road is measured using traffic flow sensors. Driving environment data includes temperature, which is detected by temperature sensors.

[0112] Furthermore, the vehicle power output data optimization method in multiple driving scenarios in this embodiment supports driving in multiple driving environments, thereby meeting the different driving needs of drivers and flexibly predicting power output data in multiple driving scenarios.

[0113] Step S302: The first sub-network predicts the power output value corresponding to the vehicle road condition data; the second sub-network combines the power output value to predict the torque demand value corresponding to traffic data and driving environment data; and the third sub-network predicts the long-term power demand value corresponding to the historical average power demand data and the energy recovery time value corresponding to the historical energy recovery time data. The first sub-network, the second sub-network, and the third sub-network constitute a multi-layer network structure.

[0114] Specifically, such as Figure 2 The diagram shown is a simplified block diagram of a multi-layer network structure. Figure 2 The network consists of a first subnetwork, a second subnetwork, and a third subnetwork, which together form a multi-layered network structure. The first subnetwork serves as the bottom layer and is used to process basic data. The second subnetwork serves as the middle layer and is used to integrate road condition data, environmental data, and traffic data. The third subnetwork serves as the top layer and is used to process historical data.

[0115] In some optional implementations, predicting a first load prediction value corresponding to vehicle traffic data through a first sub-network includes:

[0116] Step a1: Obtain speed, acceleration, and engine speed from vehicle road condition data.

[0117] Step a2: Based on speed, acceleration and engine speed, calculate the predicted power output value corresponding to the vehicle road condition data predicted by the first sub-network, wherein the predicted power output value is calculated by the following formula;

[0118] P pred (t)=f conv (v(t),a(t),n(t))

[0119] Among them, P pred (t) represents the predicted power output, f conv (v(t) is the convolution operation factor, v(t) is the velocity, a(t) is the acceleration, and n(t) is the engine speed.)

[0120] Specifically, speed, acceleration, and engine speed are extracted as features through multiple convolutional layers of the first sub-network, and then a preliminary power output prediction value is generated through a fully connected layer. This power output prediction value serves as the first prediction task.

[0121] In some optional implementations, the torque demand prediction value corresponding to vehicle road condition data, traffic data, and driving environment data is predicted by combining the power output power prediction value with the second sub-network, including:

[0122] Step b1: Obtain the slope from the vehicle road condition data, the traffic flow from the traffic data, and the temperature from the driving environment data.

[0123] Step b2: Based on the predicted power output, gradient, traffic flow, and temperature, calculate the predicted torque demand corresponding to the predicted traffic data and driving environment data of the second sub-network. The predicted torque demand is calculated using the following formula.

[0124] B pred (t)=f merge (P pred (t),θ(t),T(t),T f (t))

[0125] Among them, B pred (t) represents the predicted torque demand, f merge P is the feature fusion factor. pred (t) represents the predicted power output, θ(t) represents the slope, and T(t) represents the temperature. f (t) represents traffic flow.

[0126] Specifically, the second subnetwork combines the predicted power output value, gradient, traffic flow, and temperature output by the first subnetwork to generate the predicted torque demand value corresponding to the predicted traffic data and driving environment data of the second subnetwork.

[0127] In some optional implementations, the long-term power demand forecast corresponding to historical average power demand data and the energy recovery time forecast corresponding to historical energy recovery time data are predicted separately through a third sub-network, including:

[0128] Step c1: Based on historical average power demand data and historical energy recovery time data, calculate the long-term power demand forecast and energy recovery time forecast, which are calculated using the following formulas.

[0129] P long (t),E long (t)=f LSTM (P avg (t),E rec (t))

[0130] Among them, P long (t) represents the long-term power demand forecast, E pred (t) represents the predicted energy recovery time, f LSTM P is the computational factor for the third subnetwork. avg (t) represents the historical average power demand data, E rec (t) represents historical energy recovery time data.

[0131] Specifically, the third sub-network performs long-term power demand forecasting based on historical data. This embodiment primarily uses a Long Short-Term Memory (LSTM) network for time series modeling, with the input being the historical average power demand data P. avg (t) and historical energy recovery time data E rec (t), the output is the long-term power demand forecast P. long (t) and the predicted energy recovery time E pred (t).

[0132] The embodiments disclosed herein use multiple different sub-networks to predict various data in the current driving scenario, which helps to improve the accuracy of the prediction results of vehicle power output data.

[0133] Step S303: Based on the predicted power output, torque demand, long-term power demand, and energy recovery time, obtain the sub-network corresponding to the optimal state of the vehicle in the current driving scenario.

[0134] Step S304: Obtain the predicted value of the sub-network output corresponding to the optimal state, and use the predicted value as the optimal prediction result in the current driving scenario.

[0135] In some optional implementations, based on the predicted power output, predicted torque demand, predicted long-term power demand, and predicted energy recovery time, the sub-network corresponding to the optimal state of the vehicle in the current driving scenario is obtained, and executed using the following formula:

[0136]

[0137] Where S(t) is the state feature vector corresponding to the optimal state of the vehicle by real-time monitoring of the vehicle's current detection data in the current driving scenario, and f i (X(t),Y(t)) represents the output accuracy of the i-th sub-network corresponding to the optimal state in the current driving scenario, where X(t) is the input feature of the i-th sub-network and Y(t) is the predicted value output by the i-th sub-network.

[0138] This embodiment of the disclosure monitors real-time detection data of the vehicle in the current driving scenario (such as vehicle speed, acceleration, environmental data, and traffic data) and dynamically selects the most suitable sub-network. Compared with the traditional static model, it has higher flexibility and adaptability, and can match the prediction results of the optimal sub-network output, further improving the prediction accuracy of vehicle power output data.

[0139] In some optional implementations, the current driving scenario includes urban congestion scenarios, highway driving scenarios, or hill driving scenarios, and also includes: weighting the current detection data under urban congestion scenarios, highway driving scenarios, or hill driving scenarios.

[0140] Specifically, for urban congestion scenarios, the main focus is on handling frequent vehicle starts and stops, with an emphasis on predicting braking frequency and energy consumption at low speeds. The current detection data for urban congestion scenarios is weighted and processed using the following formula:

[0141] E city (t)=α1·v(t)+α2·a(t)+α3·f b (t)+α4·p th (t)

[0142] Where α1 is the weight corresponding to velocity, α2 is the weight corresponding to acceleration, α3 is the weight corresponding to braking frequency, α4 is the weight corresponding to braking operation threshold, v(t) is velocity, a(t) is acceleration, and f b (t) is the braking frequency, p th (t) is the braking operation threshold, E city (t) represents the energy consumption demand under urban congestion scenarios.

[0143] By weighting the current detection data in urban congestion scenarios, it becomes easier to adjust the vehicle's power output in a timely manner. For example, in urban congestion scenarios, the focus is on the vehicle's low-speed braking performance. b (t) exceeds p th If v(t) is in the low-speed range, it is determined to be an urban congestion scenario. By executing the above steps S101-S104, the predicted value of the sub-network output corresponding to the optimal state is obtained, and the predicted value is used as the optimal prediction result under the urban congestion scenario, thereby handling the frequent braking and starting under the urban congestion scenario.

[0144] Specifically, for high-speed driving scenarios, the main focus is on load prediction during stable high-speed driving, taking into account vehicle speed, engine speed, power output, and energy recovery time. The current detection data for high-speed driving scenarios is weighted and processed using the following formula:

[0145] P avg1 (t)=β1·v(t)+β2·n(t)+β3·P(t)+β4·E h (t)

[0146] Where β1 is the weight corresponding to speed, β2 is the weight corresponding to engine speed, β3 is the weight corresponding to power output, β4 is the weight corresponding to energy recovery time, v(t) is speed, n(t) is engine speed, P(t) is power output, and E h (t) represents the energy recovery time, P avg1 (t) represents the power output demand value under this high-speed driving scenario.

[0147] By weighting the current detection data in high-speed driving scenarios, it is easier to process the vehicle's power output in a timely manner. For example, when v(t) is consistently higher than a set value and n(t) meets a preset range, it is determined that the scenario is high-speed driving. In this high-speed driving scenario, by executing the above steps S301-S304, the predicted value of the sub-network output corresponding to the optimal state is obtained, and the predicted value is used as the optimal prediction result in this driving scenario, thereby executing the power output in the high-speed driving scenario.

[0148] For incline driving scenarios, the main focus is on handling situations with significant changes in slope, with a primary emphasis on adjusting vehicle torque requirements. Key considerations include vehicle speed, acceleration, and gradient. The current detection data for incline driving scenarios is weighted and processed using the following formula:

[0149] T adj (t)=γ1·θ(t)+γ2·v(t)+γ3·T(t)+γ4·a(t)

[0150] Where γ1 is the weight corresponding to the slope, γ2 is the weight corresponding to the velocity, γ3 is the weight corresponding to the torque, θ(t) is the slope, v(t) is the velocity, T(t) is the torque, a(t) is the acceleration, and T adj (t) represents the torque requirement under this slope driving scenario.

[0151] By weighting the current detection data in the slope driving scenario, it is easier to process the vehicle's power output in a timely manner. For example, when v(t) and θ(t) meet specific requirements, it is determined that the vehicle is in a slope driving scenario. In this slope driving scenario, by executing the above steps S101-S104, the predicted value of the sub-network output corresponding to the optimal state is obtained, and the predicted value is used as the optimal prediction result in this slope driving scenario, thereby executing the power output in the slope driving scenario.

[0152] This embodiment provides a method for optimizing vehicle power output data under multiple driving scenarios. It can be used with computer devices such as mobile phones, tablets, desktop computers, laptops, and servers. Step S301 involves obtaining the current detection data of the vehicle in the current driving scenario. Figure 4 As shown, the process includes the following steps:

[0153] Step S401: Acquire the data detection results of multiple sensors and the current acquisition frequency of multiple sensors of the vehicle under the current driving state within a preset time.

[0154] Specifically, the current driving state is the state corresponding to the vehicle being driven, and the preset time can be the detection time set when the sensor detection mode is turned on in the current driving state.

[0155] In a specific example, the multiple sensors include: a first sensor for detecting vehicle operation data, a second sensor for detecting vehicle road condition data, a third sensor for detecting traffic data, and a fourth sensor for detecting driving environment data.

[0156] For example, the first type of sensor includes: an acceleration sensor, a vehicle speed sensor, and a generator speed sensor. The acceleration sensor detects vehicle acceleration, and the speed sensor detects vehicle speed. The acceleration sensor transmits data to the electronic control unit via a CAN bus from the vehicle chassis, ensuring that the vehicle speed sensor synchronously acquires vehicle speed. The generator speed sensor monitors the rotational speeds of the engine and motor, reflecting this in real-time to the data monitoring system, and also monitors the current driving force and battery charge. The data detected by the first type of sensor is considered vehicle operating data. For example, acceleration... Where v(t) is the vehicle speed and t is time. The generator speed sensor collects the engine speed n(t).

[0157] For example, the second type of sensor includes a camera, a lidar sensor, and a millimeter-wave radar sensor, which detects the road slope θ(t). This second type of sensor is used to detect vehicle road condition data.

[0158] For example, the third type of sensor includes: a sensor for detecting traffic flow and a sensor for detecting traffic light data, which estimates traffic flow using the following formula: Where, N vehicles (t) represents the current road area per unit area A. lane The number of vehicles on the road. The third type of sensor is used to detect traffic data.

[0159] For example, the fourth type of sensor includes: a temperature sensor, by means of which the current ambient temperature T is detected. e (t). The fourth type of sensor is used to detect driving environment data.

[0160] This disclosure embodiment acquires data detection results of the vehicle under the current driving state through various types of sensors, which facilitates timely knowledge of the vehicle's current operating status, current road conditions, current traffic flow, and the driver's current operation.

[0161] Step S402: Based on the data detection results of the target sensor and the current acquisition frequency of the target sensor, obtain the current data change rate of the target sensor; wherein, the target sensor is any one of a variety of sensors.

[0162] Specifically, for example, the target sensor is the first sensor among multiple sensors. The current acquisition frequency of the first sensor is 5 acquisitions every 10 minutes. Based on the data detection results of the first sensor in 5 acquisitions, the current data change rate of the first sensor is obtained. For example, the first sensor is specifically a vehicle speed sensor. The vehicle speed detected in the first acquisition is 30 km / h, the second is 31 km / h, the third is 30 km / h, the fourth is 33 km / h, and the fifth is 32 km / h. The change between the first and second acquisitions is 1, the change between the second and third is 1, the change between the third and fourth is 3, and the change between the fourth and fifth is 1. The current data change rate of the target sensor is ΔC. s (t) = (1 + 1 + 3 + 1) / 4 = 1.5.

[0163] Step S403: Calculate the dynamic adjustment result of the target sensor data based on the current acquisition frequency and the current data change rate of the target sensor.

[0164] Step S404: Based on the data dynamic adjustment results, adjust the current acquisition frequency of the target sensor.

[0165] Specifically, during vehicle operation, sudden environmental changes can occur, such as sudden rain, hill starts, or highway driving. Taking a sudden rainstorm as an example, if sensors collecting ambient humidity data at a fixed frequency fail to adapt to changes in the external environment, the drive system cannot adjust power output or other critical control parameters in a timely manner. Conversely, if sensors continuously collect data at a fixed frequency under stable driving conditions, the large amount of similar data collected leads to data redundancy and increases the computational burden on the drive system, ultimately affecting its overall efficiency. Furthermore, when multiple sensors work simultaneously, collecting data from all sensors at a fixed frequency increases system latency and resource consumption. Therefore, it is necessary to flexibly adjust the current sampling frequency of the target sensors based on actual driving conditions.

[0166] In some optional implementations, the dynamic adjustment result of the target sensor data is calculated using the following formula;

[0167]

[0168] Among them, f sensor (t) represents the current sampling frequency of the target sensor, ΔC s (t) represents the rate of change of the current data from the target sensor, ΔC th f is a preset threshold. high f is the first sampling frequency. low The second sampling frequency is the first sampling frequency, which can be a high-frequency frequency, and the second sampling frequency can be a low-frequency frequency. The target sensor can be any of a variety of sensors.

[0169] For example, in heavy rain, the road surface friction coefficient affects vehicle driving performance. Therefore, the data from the friction coefficient sensor, one of the fourth sensors that detects driving environment data, will have a higher priority, typically ΔC. s (t)≥ΔC th According to f sensor (t)=f high This ensures data is collected at a high frequency. However, when air humidity has a relatively small impact on the vehicle, the data priority of the humidity sensor (the fourth type of sensor that detects driving environment data) will be reduced, ΔC. s (t)<ΔC th According to f sensor =flow .

[0170] Ensure that data is collected at low frequencies to save system resources and avoid redundant transmission of irrelevant data.

[0171] For example, when a vehicle travels at a constant speed on a flat road, continuously collecting data at a specific high frequency would waste data resources. In this case, the data priority of the speed sensor, the first type of sensor used to detect vehicle operation data, would be reduced, ΔC s (t)<ΔC th According to f sensor =f low This ensures that data is collected at low frequencies, thereby saving system resources and avoiding redundant transmission of irrelevant data.

[0172] This embodiment provides a method for optimizing vehicle power output data under multiple driving scenarios. It can be used with computer devices such as mobile phones, tablets, desktop computers, laptops, and servers. After step S401, which involves acquiring data detection results from multiple sensors and their current acquisition frequencies within a preset time period, the following steps are performed: Figure 5 As shown, the process also includes the following steps:

[0173] Step S501: Calculate the multi-sensor collaborative detection result based on the data detection results of multiple sensors and the weight parameters of multiple sensors.

[0174] In some optional implementations, the multi-sensor collaborative detection results are calculated using the following formula;

[0175]

[0176] Among them, O sync (t) represents the multi-sensor collaborative detection parameters, w i Let R be the weighting parameter for the i-th sensor. i (t) represents the detection result of the i-th sensor, and n represents the total number of various sensors.

[0177] When a vehicle is in motion, if the data detection result of a certain sensor remains stable, the data detection result of that sensor can be shared with other sensors, thereby avoiding the repeated collection of related data.

[0178] Step S502: Based on the data dynamic adjustment results and the multi-sensor collaborative detection results, obtain the data sharing results of multiple sensors.

[0179] In a specific example, step S402 above, based on the dynamic data adjustment results and the multi-sensor collaborative detection results, obtains data sharing results from multiple sensors, including:

[0180] Step d1: Based on the data dynamic adjustment results, obtain the current acquisition frequency of the target sensor after dynamic adjustment.

[0181] Specifically, for example, in heavy rain, the road surface friction coefficient affects the vehicle's driving performance. Therefore, the data from the friction coefficient sensor, one of the fourth sensors used to detect driving environment data, will have a higher priority, typically ΔC. s (t)≥ΔC th According to f sensor (t)=f high To ensure data is collected at a high frequency, the dynamic adjustment result of the target sensor data is f. sensor (t)=f high The current acquisition frequency of the target sensor after dynamic adjustment is f. high However, when air humidity has a relatively small impact on the vehicle, the data priority of the humidity sensor, one of the fourth sensors used to detect driving environment data, will be reduced, ΔC. s (t)<ΔC th According to f sensor =f low At this point, the dynamic adjustment result of the target sensor data is f. sensor (t)=f low .

[0182] Step d2: If the current acquisition frequency of the target sensor after dynamic adjustment remains stable within a preset time, the data detection result of the target sensor after multi-sensor collaboration is obtained in the stable state.

[0183] Step d3 involves using the data detection results of the target sensor after multi-sensor collaboration as the data sharing result of multiple sensors.

[0184] Specifically, for example, during the uphill driving of a vehicle, the camera, lidar sensor, and millimeter-wave radar sensor in the second type of sensor are used to acquire road visual information. Since the camera, lidar sensor, and millimeter-wave sensor can all be used to detect road visual information, when the current acquisition frequency of the target sensor after dynamic adjustment of the camera remains stable within a preset time, the data detection result of the camera can be used as the data sharing result of multiple sensors.

[0185] This disclosure, by dynamically adjusting the data from the target sensor, facilitates flexible adjustment of the sensor's current acquisition frequency, enabling timely adaptation to rapidly changing external environments. This, in turn, allows the drive system to adjust power output or other key control parameters promptly. Furthermore, it avoids data redundancy and increased computational burden on the drive system when collecting large amounts of similar data under stable vehicle operating conditions, significantly improving the overall efficiency of the drive system. In addition, this disclosure, by combining the dynamic data adjustment results with the calculation of multi-sensor collaborative detection results, facilitates collaborative computing and information sharing among multiple sensors, thereby improving the overall system efficiency and the effectiveness of data processing. Ultimately, this enables the drive system to make efficient and intelligent responses in complex environments.

[0186] like Figure 6 The diagram shown is a simplified block diagram illustrating real-time vehicle detection data collection in multiple driving scenarios according to an embodiment of this disclosure. Figure 6 In this system, the vehicle collects real-time data from various sensors while in its current driving state. This data primarily includes five categories: vehicle operation data, road condition data, traffic data, and driving environment data. These five types of data undergo preprocessing and optimization. Following optimization, based on dynamic data adjustments and multi-sensor collaborative detection results, shared data from multiple sensors is obtained.

[0187] This embodiment provides a method for optimizing vehicle power output data under multiple driving scenarios, which can be used in computer devices such as mobile phones, tablets, desktop computers, laptops, servers, etc. Figure 7 As shown, the process also includes the following steps:

[0188] Step S701: Preprocess the data detection results from multiple sensors.

[0189] In some optional implementations, the data detection results from multiple sensors are preprocessed, including:

[0190] Step b1 involves preprocessing the data detection results from multiple sensors.

[0191] In a specific example, the preprocessing methods include outlier handling and missing value imputation.

[0192] Specifically, due to environmental interference or hardware errors during vehicle operation, sensor data may show unreasonable outliers. For example, temperature sensors may exhibit sudden changes in stable environments; such abnormal data needs to be identified and filtered.

[0193] In a specific example, outlier handling is done by using the mean-standard deviation method to identify outliers, which is expressed by the following formula:

[0194]

[0195] When Z i (t) exceeds the predetermined threshold (|Z) i If (t)|>3), then the data is considered an outlier and needs to be processed.

[0196] In a specific example, the missing value imputation method is to use linear interpolation to fill in the missing values, which is expressed by the following formula;

[0197]

[0198] Among them, X i (t prev X represents the normal data points before the missing values. i (t next ) represents the normal data points after the missing values, t prev t represents the time point before the missing values. next t represents the time point following the missing value, and t represents the current time point.

[0199] Step b2 involves processing the preprocessed data detection results from multiple sensors again using an adaptive Kalman filter algorithm.

[0200] In a specific example, the adaptive Kalman filter algorithm is executed using the following formula:

[0201]

[0202] K(t)=P(t∣t-1)H T (HP(t∣t-1)H T +R(t)) -1

[0203]

[0204] in, Let F be the state estimation matrix of the i-th sensor at time t-1, F be the state transition matrix of the i-th sensor at time t-1, Bu(t) be the control input matrix at time t, K(t) be the Kalman gain matrix at time t, P be the covariance matrix of the i-th sensor at time t-1, and H be the state estimation matrix of the i-th sensor at time t-1. T Let R(t) be the observation matrix, and R(t) be the observation noise covariance matrix. Let X be the state estimation matrix of the i-th sensor at time t. i (t) represents the prediction result of the i-th sensor at time t.

[0205] Specifically, in the above formula, the state estimation matrix for the next time step is predicted using the state transition matrix F and the control input matrix Bu(t), and further based on the error covariance matrix P and the observation matrix H... T Observe the noise covariance matrix R(t), calculate the Kalman gain K(t), and use the Kalman gain K(t) to update the current state estimation matrix. To make the data more closely resemble actual observations, this embodiment of the disclosure uses an adaptive Kalman filter algorithm to adjust the Kalman gain in real time, thereby dynamically adjusting the sensor's data detection results. The error covariance matrix is ​​used to estimate the difference between the current observation and the predicted value, and the sensor's data detection results are corrected based on this difference, making the updated estimate closer to the actual data, ultimately achieving a data optimization prediction effect. For example, when the sensor noise is low or the data is relatively stable, the Kalman gain K(t) will decrease, reducing the sensor's current amplitude to adapt to the current changes; conversely, when the noise is high or the data fluctuations are significant, the Kalman gain K(t) will increase, reducing the sensor's current amplitude to adapt to the current changes.

[0206] Step S702: Obtain acceleration and velocity from the data detection results of multiple sensors.

[0207] Step S703: Create physical constraints for acceleration and velocity using the following formula.

[0208] v min ≤v(t)≤v max

[0209] a min ≤a(t)≤a max

[0210] Step S704: Based on the physical constraints of acceleration and velocity, the preset optimization algorithm is expressed by the following formula.

[0211]

[0212] Among them, V min V is the minimum speed value. max V is the first speed threshold. max Let v(t) be the second velocity threshold, v(t) be the velocity at time t, and a be the velocity threshold. min a is the first acceleration threshold. max The second acceleration threshold, v(t) is the acceleration at time t;

[0213] Step S705: Optimize the filtered data detection results of multiple sensors according to the preset optimization algorithm.

[0214] Specifically, since speed and acceleration are directly related to a vehicle's dynamic performance during operation, physical constraints on acceleration and speed are created. A pre-defined optimization algorithm ensures that the filtered data better reflects the vehicle's actual operating conditions, thereby improving the accuracy of data detection results from multiple sensors.

[0215] This embodiment also provides a power output data optimization device for a vehicle under multiple driving scenarios. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0216] This embodiment provides a device for optimizing vehicle power output data under multiple driving scenarios, such as... Figure 8 As shown, it includes:

[0217] The first acquisition module 801 is used to acquire, in the current driving scenario, the predicted power output value based on vehicle road condition data, the predicted torque demand value based on traffic data and driving environment data, and the predicted energy recovery time value based on historical energy recovery time data.

[0218] The second acquisition module 802 is used to acquire the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario.

[0219] The third acquisition module 803 is used to acquire the first execution parameter corresponding to the predicted value of power output power based on the actual value of power output power, acquire the second execution parameter corresponding to the predicted value of torque demand based on the actual value of torque demand, and acquire the third execution parameter corresponding to the predicted value of energy recovery time based on the actual value of torque demand.

[0220] The data prediction module 804 is used to predict the power output power prediction value through the target neural network, generate the power output power optimization value through the first execution parameter, predict the torque demand prediction value through the target neural network, generate the torque demand optimization value through the second execution parameter, predict the energy recovery time prediction value through the target neural network, generate the energy recovery time optimization value through the third execution parameter, and generate a loss function between the optimization value of each of the above items and the actual value corresponding to the optimization value. The loss function is used to update the weight parameter values ​​of the target neural network.

[0221] The data optimization module 805 is used to collaboratively optimize the power output, torque demand, and energy recovery time values ​​within the driving cycle under the current driving scenario using global and local optimization algorithms, respectively.

[0222] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0223] In this embodiment, the vehicle power output data optimization device under multiple driving scenarios is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0224] This invention also provides a computer device having the power output data optimization device described above.

[0225] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 9 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 9 Take a processor 10 as an example.

[0226] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0227] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0228] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0229] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0230] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0231] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that a computer, processor, microprocessor controller, or programmable hardware includes storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented. Although embodiments of the invention have been described with reference to the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for optimizing vehicle power output data under multiple driving scenarios, characterized in that, The method includes: In the current driving scenario, the predicted power output based on vehicle road condition data, the predicted torque demand based on traffic data and driving environment data, and the predicted energy recovery time based on historical energy recovery time data are obtained respectively. Obtain the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario; Based on the actual power output value, a first execution parameter corresponding to the predicted power output value is obtained; based on the actual torque demand value, a second execution parameter corresponding to the predicted torque demand value is obtained; and based on the actual torque demand value, a third execution parameter corresponding to the predicted energy recovery time value is obtained. The target neural network predicts the power output power prediction value and generates the power output power optimization value using the first execution parameter. The target neural network also predicts the torque demand prediction value and generates the torque demand optimization value using the second execution parameter. The target neural network predicts the energy recovery time prediction value and generates the energy recovery time optimization value using the third execution parameter. A loss function is generated between the optimization value of each of the above items and the actual value corresponding to the optimization value. The loss function is used to update the weight parameter values ​​of the target neural network. The power output, torque demand, and energy recovery time are optimized collaboratively using global and local optimization algorithms within the driving cycle of the current driving scenario.

2. The method according to claim 1, characterized in that, Based on the actual power output value, obtain the first execution parameter corresponding to the predicted power output value, and execute it using the following formula: ΔP(t)=λ P ·(P actual (t)-P pred (t)) P exec (t)=P pred (t)+ΔP(t) Where ΔP(t) is the first deviation between the actual value of the power output power and the predicted value of the power output power, and λ P P is the first weighting parameter. actual (t) represents the actual value of the power output, P pred (t) represents the predicted power output value, P exec (t) is the first execution parameter corresponding to the predicted power output value.

3. The method according to claim 1, characterized in that, Based on the actual torque demand value, obtain the second execution parameter corresponding to the predicted torque demand value, and execute it using the following formula: ΔB(t)=λ B ·(B factual (t)-B pred (t)) B exec (t)=B pred (t)+ΔB(t) Where ΔB(t) is the second deviation between the actual torque demand and the predicted torque demand, and λ B B is the second weighting parameter. actual (t) represents the actual value of the torque requirement, B pred (t) represents the predicted torque demand value, B pred (t) represents the predicted torque demand value, B exec (t) is the second execution parameter.

4. The method according to claim 1, characterized in that, The predicted energy recovery time, based on historical energy recovery time data, is calculated using the following formula: ΔE(t)=λ E ·(E actual (t)-E pred (t)) E exec (t)=E pred (t)+ΔE(t) Where ΔE(t) is the third deviation between the actual value of the energy recovery time and the predicted value of the energy recovery time, λ E E is the third weighting parameter. actual (t) represents the actual value of the energy recovery time, E pred (t) is the predicted energy recovery time, E exec (t) is the third execution parameter.

5. The method according to claim 1, characterized in that, A loss function is generated between the optimized value and the corresponding actual value for each of the above terms. This loss function is used to update the weight parameter values ​​of the target neural network and is executed using the following formula: in, Let the loss function be... For each of the above predicted values, y i The actual value corresponding to each of the above optimized values, Ω(f) k ) represents the regularization term. For the error loss term, f k The target neural network uses the Kth tree of a decision tree, where m is the number of samples used to train the target neural network, M is the total number of target neural networks, and w h (t) represents the weighting parameter for historical vehicle detection data, w r (t) represents the weighting parameter of the current vehicle detection data. This is a predicted value based on historical data. The predicted value is based on the current data. For each of the above items, the optimized value is given.

6. The method according to claim 1, characterized in that, The power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a global optimization algorithm, and are executed using the following formula: Where, r global,1 (t) is the reward function of the global optimization algorithm, A1 is the set of actions to be executed by the global optimization algorithm, and π global,1 (t) represents the optimal result corresponding to the optimal action taken by the global optimization algorithm.

7. The method according to claim 1, characterized in that, The power output, torque demand, and energy recovery time are further optimized within the driving cycle of the current driving scenario using a local optimization algorithm, employing the following formula: r local,2 (t)=-|M actual (t)-M pred (t)| Where, r local,2 (t) is the reward function of the local optimization algorithm, A2 is the set of actions to be performed by the local optimization algorithm, and π local,2 (t) represents the optimal result corresponding to the optimal action taken by the local optimization algorithm, M. actual (t) represents the actual value described in each of the above items, M pred (t) represents the predicted value for each of the above items.

8. The method according to claim 1, characterized in that It also includes: the optimization results of the local optimization algorithm and the global optimization algorithm are combined using the following formula; Where, π i (t) represents the weight of the optimization result of the combined global optimization algorithm and the local optimization algorithm in the current driving scenario, r scene,i (t) represents the reward value of the optimization result of the combined global optimization algorithm and the local optimization algorithm in the current driving scenario, r global,3 (t) represents the final result after collaboration, and N represents the number of collaborating objects.

9. A device for optimizing vehicle power output data under multiple driving scenarios, characterized in that, The device includes: The first acquisition module is used to acquire, in the current driving scenario, the predicted power output value based on vehicle road condition data, the predicted torque demand value based on traffic data and driving environment data, and the predicted energy recovery time value based on historical energy recovery time data. The second acquisition module is used to acquire the actual values ​​of the vehicle's power output, torque demand, and energy recovery time under the current driving scenario. The third acquisition module is used to acquire, based on the actual value of the power output power, a first execution parameter corresponding to the predicted value of the power output power, based on the actual value of the torque demand, a second execution parameter corresponding to the predicted value of the torque demand, and based on the actual value of the torque demand, a third execution parameter corresponding to the predicted value of the energy recovery time. The data prediction module is used to predict the power output power prediction value through the target neural network, generate the power output power optimization value through the first execution parameter, predict the torque demand prediction value through the target neural network, generate the torque demand optimization value through the second execution parameter, predict the energy recovery time prediction value through the target neural network, generate the energy recovery time optimization value through the third execution parameter, and generate a loss function between the optimization value and the actual value corresponding to each of the above items. The loss function is used to update the weight parameter values ​​of the target neural network. The data optimization module is used to further optimize the power output value, torque demand value, and energy recovery time value within the driving cycle under the current driving scenario using both global and local optimization algorithms.

10. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the power output data optimization method for a vehicle in multiple driving scenarios as described in any one of claims 1 to 8.

Images