Vehicle control device

Abstract

The invention relates to a control device for a vehicle. The present invention addresses the problem of providing a control device for a vehicle that can use a learned model to give consideration to both the protection of a drive system and the trafficability. Provided is a vehicle control device. The apparatus includes a prediction unit that predicts a feature amount by inputting travel data of a vehicle into a first learned model obtained by machine learning, and an estimation unit that estimates whether the vehicle can pass through a road surface by inputting the feature amount into a second learned model obtained by machine learning. A setting unit that determines a limit value for limiting the driving force or braking force of the vehicle on the basis of the estimation by the estimation unit; and a control unit that controls the limit value on the basis of the estimation and the result of whether the vehicle can pass.

Description

Technical Field

[0001] This invention relates to a vehicle control device. Background Technology

[0002] A technique has been developed to predict driving operations using a fully trained model of a neural network. By predicting driving operations using a fully trained model, excessive load on transmissions and other components can be avoided (e.g., Patent Document 1).

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application Publication No. 2022-174435 Summary of the Invention

[0006] The problem that the invention aims to solve

[0007] In vehicles, it is necessary to protect the drive system while ensuring passability. Therefore, the objective is to provide a vehicle control device that can use a learned model to balance drive system protection and passability.

[0008] Technical solutions for solving the problem

[0009] The above objective can be achieved by a vehicle control device comprising: a prediction unit that inputs vehicle driving data into a first learned model obtained through machine learning to predict feature quantities; a prediction unit that inputs the feature quantities into a second learned model obtained through machine learning to predict whether the vehicle can pass through the road surface; a setting unit that determines a limit value for limiting the driving force or braking force of the vehicle based on the prediction of the prediction unit; and a control unit that controls the limit value based on the prediction and the result of whether the vehicle can pass through the road surface.

[0010] Invention Effects

[0011] It can provide a vehicle control device that can use a learned model to balance the protection of the drive system and the passability. Attached Figure Description

[0012] Figure 1 This is a schematic structural diagram of the vehicle according to this embodiment.

[0013] Figure 2 This is a flowchart illustrating the process in the implementation method.

[0014] Figure 3 This is a flowchart illustrating the process in the implementation method.

[0015] (Symbol Explanation)

[0016] 1: Vehicle; 10: Internal combustion engine; 12: Automatic transmission; 13: Transfer case; 16: Drive wheel; 20: Drive shaft; 22: Differential gear; 24: Drive shaft; 26: Accelerator pedal; 28: Accelerator opening sensor; 30: Wheel speed sensor; 32: Braking mechanism; 40: ECU; 42: Prediction unit; 44: Speculation unit; 46: Setting unit; 48: Control unit. Detailed Implementation

[0017] The vehicle control device of this embodiment will now be described with reference to the accompanying drawings. Figure 1 This is a schematic structural diagram of vehicle 1 according to this embodiment. Vehicle 1 is a four-wheel drive vehicle. Vehicle 1 includes an internal combustion engine 10, an automatic transmission 12, a transfer case 13, four drive wheels 16, two drive shafts 20, two differential gears 22, a drive shaft 24, an accelerator pedal 26, an accelerator opening sensor 28, wheel speed sensors 30, a braking mechanism 32, and an ECU (Electronic Control Unit) 40.

[0018] The internal combustion engine 10 is, for example, a gasoline engine or a diesel engine. The automatic transmission 12 is a stepped transmission or a continuously variable transmission. The automatic transmission 12 is connected to the internal combustion engine 10. The transfer case 13 is connected to the automatic transmission 12.

[0019] Two drive shafts 20 are connected to the transfer case 13. One of the two drive shafts 20 extends forward of the vehicle 1. The other drive shaft 20 extends rearward of the vehicle 1. The two drive shafts 20 are respectively connected to the differential gear 22. The drive shaft 24 is connected to the differential gear 22 and the drive wheel 16.

[0020] Vehicle 1 is driven by the driver operating the accelerator pedal 26. The internal combustion engine 10 is the power source of vehicle 1, generating driving force by burning fuel such as gasoline. The driving force output from the internal combustion engine 10 is transmitted to the drive wheels 16 via the automatic transmission 12, transfer case 13, drive shaft 20, and drive shaft 24. The drive wheels 16 rotate due to the driving force, and vehicle 1 moves.

[0021] The greater the amount of time the accelerator pedal 26 is depressed (accelerator opening), the greater the increase in the torque output (engine torque) of the internal combustion engine 10. The smaller the accelerator opening, the less torque is produced. The accelerator opening sensor 28 detects the accelerator opening.

[0022] The number of wheel speed sensors 30 and braking mechanisms 32 is the same as that of the drive wheels 16, which is four. Wheel speed sensors 30 and braking mechanisms 32 are respectively installed at each of the four drive wheels 16. The wheel speed sensors 30 detect the rotational speed (wheel speed) of the drive wheels 16. The braking mechanism 32 is, for example, a hydraulic brake, which applies braking force to the drive wheels 16, thereby slowing down the rotation of the drive wheels 16.

[0023] ECU 40 is the control device of vehicle 1, equipped with a CPU (Central Processing Unit) and other computing devices, RAM (Random Access Memory) and ROM (Read Only Memory) and other storage devices. ECU 40 performs various controls by executing programs stored in ROM and other storage devices. ECU 40 obtains the accelerometer opening from accelerometer opening sensor 28. ECU 40 obtains the wheel speeds of the four drive wheels 16 from the four wheel speed sensors 30.

[0024] The ECU 40 controls the four braking mechanisms 32 based on factors such as the amount of brake pedal depressed (not shown), thereby increasing or decreasing the braking force. The ECU 40 also controls the torque of the internal combustion engine 10 based on the accelerator pedal opening.

[0025] Depending on the road surface and driving conditions, the torque applied to the drive system, including the drive shaft 20, differential gear 22, and drive shaft 24, varies. For example, the ECU 40 performs traction control (TRC), adjusting the engine torque and the braking force of the braking mechanism 32 to prevent the drive wheels 16 from spinning freely. The torque is applied to the drive system based on the braking force. Additionally, when a difference in speed occurs between the left and right drive wheels 16, a torque limiting the differential is generated from the differential gear 22. Excessive torque applied to the drive system poses a risk of damage. Higher torque output from the internal combustion engine 10 and higher braking force from the braking mechanism 32 result in a greater torque being applied to the drive system. Lower engine torque and lower braking force result in lower torque in the drive system, but also a decrease in the vehicle's passability. In this embodiment, by controlling the limiting values ​​of engine torque or braking force, the torque of the drive system is controlled, and passability is improved.

[0026] The ECU40's storage device stores AI (Artificial Intelligence) models, including a first fully learned model and a second fully learned model. These models are created beforehand using experimental and simulation data as training data and through machine learning. The first fully learned model is created, for example, using LSTM (Long Short-Term Memory). The second fully learned model is created, for example, using Random Forest. The first fully learned model predicts feature quantities based on the input driving data of vehicle 1. Feature quantities refer to the future wheel speed, body speed, torque, etc. of vehicle 1. The second fully learned model infers whether vehicle 1 can pass the road surface based on the predicted feature quantities.

[0027] ECU 40 functions as a prediction unit 42 that inputs driving data into the first learned model to predict feature quantities. ECU 40 functions as a prediction unit 44 that inputs feature quantities into the second learned model to predict whether vehicle 1 can pass the road surface. ECU 40 functions as a setting unit 46 that sets a limit value for at least one of engine torque and braking force based on the prediction to limit them. ECU 40 functions as a control unit 48 that controls the limit value based on the result of whether vehicle 1 can pass the road surface.

[0028] Figure 2 and Figure 3 This is a flowchart illustrating the process in the embodiment. ECU 40 acquires data (step S10). The data includes driving data and road surface data. Driving data refers to accelerator opening, engine torque, internal combustion engine speed 10, automatic transmission gear 12 gear position, wheel speed, hydraulic pressure of braking mechanism 32, vehicle 1 acceleration, and information about the surroundings of vehicle 1, etc. Road surface data refers to data related to road surface types such as paved roads and muddy roads. Road surface data can be acquired from an onboard camera, or it can be inferred from driving data.

[0029] Based on road surface data, ECU40 selects a first learned model suitable for the road surface from the stored machine learning learned models. ECU40 inputs driving data into the selected first learned model. The first learned model predicts feature values ​​(step S12). ECU40 inputs the feature values ​​into a second learned model to infer whether vehicle 1 can pass the road surface (whether it can pass or gets stuck) (step S14).

[0030] If a positive determination is made in step S14 ("Yes", indicating that passage is possible), ECU40 selects the value L1 corresponding to the predicted passage as the limit value for engine torque (step S16). ECU40 then determines whether vehicle 1 can actually pass (step S18). If a positive determination is made (passable), the process ends.

[0031] If a negative judgment ("No", cannot pass) is made in step S18, ECU40 selects the value L2, which is the value when it cannot pass (trapped), as the limit value of engine torque (step S20). ECU40 predicts the maximum value Tr of the torque applied to the drive system based on engine torque, gear ratio, brake force, ratio of left and right wheel speeds, inertia, etc. ECU40 determines whether the maximum torque Tr is lower than the threshold Tth (step S22).

[0032] If a positive determination is made, ECU40 determines whether the engine torque has reached the limit value L2 (step S24). If a negative determination is made in either step S22 or S24, the process ends. If a positive determination is made in both steps S22 and S24, ECU40 changes the trapped limit value L2 to a value larger than the value selected in step S20 (step S26). ECU40 sets a relearning flag for the second learned model (step S28). The second learned model, which yielded an incorrect result in the prediction of whether it could pass, is relearned. After step S28, the process ends.

[0033] If a negative determination is made in step S14 ("No", presumed to be stuck), ECU40 selects the value L2 corresponding to the presumed (cannot pass) as the limit value for engine torque (step S30). ECU40 determines whether vehicle 1 has actually gotten stuck (step S32). If a positive determination is made (stuck), the process ends.

[0034] If a negative determination ("No, can pass") is made in step S32, ECU40 selects the passable value L1 as the engine torque limit value (step S34). ECU40 determines whether the maximum torque Tr applied to the drive system is above the threshold Tth (step S36). If a positive determination is made, ECU40 determines whether the engine torque reaches the limit value L1 (step S38). If a negative determination is made in either step S36 or S38, the process ends. If a positive determination is made in both step S36 and S38, ECU40 changes the passable limit value L1 to a value smaller than the value selected in step S24 (step S40). ECU40 sets the relearning flag for the second learned model (step S42). After step S42, the process ends.

[0035] According to the implementation method, feature quantities are predicted by inputting driving data into a first fully learned model. The feature quantities are then input into a second fully learned model to infer whether vehicle 1 can traverse the road surface. If the driving result matches the prediction, the limit value remains unchanged. If the driving result differs from the prediction, ECU 40 changes the limit value.

[0036] Specifically, if it is predicted that the vehicle can pass but gets stuck, the engine torque limit value L2 when stuck is increased. This improves passability. If it is predicted that the vehicle is stuck but passes, the passability limit value L1 is decreased. This reduces the torque applied to the drive system, protecting it. This balances drive system protection and passability.

[0037] In the above example, the engine torque is controlled by a limit value, but it is also possible to control one or both of the engine torque and the braking force. By increasing the limit values ​​for both engine torque and braking force, throughput is improved. Figure 2 Step S26). By lowering the limit value, passability is improved ( Figure 3 (Step S40). When the vehicle 1 has an electric generator (MG) as a power source, the ECU 40 determines a limit value for the motor torque and controls the limit value. For example, a limit value can also be determined for the speed of the internal combustion engine 10.

[0038] By setting the relearning flag to ON (steps S28 and S42), the model, which has completed its second learning stage, is relearned. For example, by inputting data that yields an incorrect result for passing or failing, the model is relearned so that it can yield the correct result for passing or failing. Through relearning, the accuracy of passing or failing determination is improved.

[0039] The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the specific embodiments described above. Various modifications and alterations can be made within the scope of the spirit of the present invention as set forth in the claims.

Claims

1. A control device of a vehicle, comprising: a prediction unit configured to input travel data of the vehicle to a first learned model obtained by learning through machine learning to predict a characteristic quantity; a presumption unit configured to input the characteristic quantity to a second learned model obtained by learning through machine learning to presume whether the vehicle can pass a road surface; a setting unit configured to determine a limit value that limits a driving force or a braking force of the vehicle based on the presumption of the presumption unit; and a control unit configured to control the limit value based on the presumption and a result of whether the vehicle can pass. ​

Images