Control device of vehicle drive device, vehicle-mounted electronic control unit, trained model, machine learning system, method of controlling vehicle drive device, method of producing electronic control unit, and output parameter calculation device

Patent metadata

Field	Value
Publication number	US-10864900-B2
Application number	US-201916569088-A
Country	US
Kind code	B2
Filing date	Sep 12, 2019
Priority date	Oct 9, 2018
Publication date	Dec 15, 2020
Grant date	Dec 15, 2020

Abstract

A control device 60 of a vehicle drive device comprises a processing part 81 configured to use a trained model using a neural network to calculate at least one output parameter of a vehicle, and a control part 82 configured to control the vehicle drive device. The neural network includes a first input layer to which at least one first input parameter of the vehicle at a first point of time is input, a second input layer to which at least one second input parameter of the vehicle at a second point of time is input, a first hidden layer to which outputs of the first input layer are input, a second hidden layer to which at least one value correlated with the outputs of the first hidden layer, and outputs of the second input layer are input, and an output layer outputting at least one output parameter.

First claim

The invention claimed is:

1. A control device of a vehicle drive device comprising
a processing part configured to use a trained model using a neural network to calculate at least one output parameter of a vehicle, and
a control part configured to control the vehicle drive device mounted in the vehicle based on the at least one output parameter calculated by the processing part, wherein
the neural network includes a first input layer to which at least one first input parameter of the vehicle at a first point of time is input, a second input layer to which at least one second input parameter of the vehicle at a second point of time after the first point of time is input, a first hidden layer to which outputs of the first input layer are input, a second hidden layer to which at least one value correlated with the outputs of the first hidden layer, and outputs of the second input layer are input, and an output layer outputting at least one output parameter.

2. The control device of a vehicle drive device according to claim 1 , wherein the number of the second input parameters is smaller than the number of the first input parameters.

3. The control device of a vehicle drive device according to claim 2 , wherein the second input parameters are a part of the first input parameters.

4. The control device of a vehicle drive device according to claim 1 , wherein the neural network is configured so that outputs of the first hidden layer are input to the second hidden layer.

5. The control device of a vehicle drive device according to claim 1 , wherein the neural network includes at least one hidden layer between the first hidden layer and the second hidden layer.

6. The control device of a vehicle drive device according to claim 1 , wherein the neural network is configured so that outputs of the second hidden layer are input to the output layer.

7. The control device of a vehicle drive device according to claim 1 , wherein the neural network includes at least one hidden layer between the second hidden layer and the output layer.

8. The control device of a vehicle drive device according to claim 1 , wherein
the control part is configured to control an internal combustion engine mounted in the vehicle, and
the first input parameters include at least one of engine speed, engine water temperature, concentration of oxygen in intake inside an intake manifold, cylinder intake amount, cylinder volume, cylinder pressure, cylinder temperature, fuel injection amount, fuel injection timing, and fuel injection pressure, the second input parameters include at least one of a fuel injection amount, fuel injection timing, and fuel injection pressure, and the output parameters include at least one of a sound pressure of combustion noise, a concentration of harmful substances discharged from an engine body, and a heat efficiency of the internal combustion engine.

9. The control device of a vehicle drive device according to claim 8 , wherein the first input parameters include the cylinder pressure and cylinder temperature, and the second input parameters do not include the cylinder pressure and cylinder temperature.

10. A vehicle-mounted electronic control unit comprising a processing part configured to receive a trained model using a neural network through a communication device provided at a vehicle from a server at an outside of the vehicle and using the trained model to calculate at least one output parameter of the vehicle, wherein
the neural network includes a first input layer to which at least one first input parameter of the vehicle at a first point of time is input, a second input layer to which at least one second input parameter of the vehicle at a second point of time after the first point of time is input, a first hidden layer to which outputs of the first input layer are input, a second hidden layer to which at least one value correlated with the outputs of the first hidden layers, and outputs of the second input layer are input, and an output layer outputting at least one output parameter, and
the server comprises a storage device storing sets of training data including the first input parameters, the second input parameters, and the output parameters and is configured to use the sets of training data to generate the trained model.

11. A vehicle-mounted electronic control unit comprising
a parameter acquiring part configured to acquire at least one first input parameter of a vehicle at a first point of time, at least one second input parameter of the vehicle at a second point of time after the first point of time, and at least one output parameter of the vehicle at a point of time after the second point of time, and send the at least one first input parameter, the at least one second input parameter, and at least one output parameter through a communication device provided at the vehicle to a server at an outside of the vehicle, and
a processing part configured to receive a trained model using a neural network from the server through the communication device and using the trained model to calculate the at least one output parameter, wherein
the neural network includes a first input layer to which the at least one first input parameter is input, a second input layer to which the at least one second input parameter is input, a first hidden layer to which outputs of the first input layer are input, a second hidden layer to which at least one value correlated with the outputs of the first hidden layer, and outputs of the second input layer are input, and an output layer outputting the at least one output parameter, and
the server uses the at least one first input parameter, the at least second input parameter, and the at least output parameters acquired by the parameter acquiring part as sets of training data to generate the trained model.

12. A trained model using a neural network including a first input layer to which at least one first input parameter of a vehicle at a first point of time is input, a second input layer to which at least one second input parameter of the vehicle at a second point of time after the first point of time is input, a first hidden layer to which outputs of the first input layer are input, a second hidden layer to which at least one value correlated with the outputs of the first hidden layer, and outputs of the second input layer are input, and an output layer outputting at least one output parameter of the vehicle, wherein
weights of the neural network have been trained using the at least one first input parameter, the at least one second input parameter, and the at least one output parameter as sets of training data.

13. A machine learning system comprising an electronic control unit provided at a vehicle, a communication device provided at the vehicle, and a server at an outside of the vehicle, wherein,
the electronic control unit comprises
a parameter acquiring part configured to acquire at least one first input parameter of the vehicle at a first point of time, at least one second input parameter of the vehicle at a second point of time after the first point of time, and at least one output parameter of the vehicle at a point of time after the second point of time, and send the at least one first input parameter, the at least one second input parameter, and the at least one output parameter through the communication device to the server, and
a processing part configured to receive a trained model using a neural network from the server through the communication device and use the trained model to calculate the at least one output parameter, wherein
the neural network includes a first input layer to which the at least one first input paramete

Assignees

• Toyota Motor Co Ltd

Inventors

• Nakamura Toshihiro

Classifications

• G06N3/048

  Activation functions · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• F02D41/2477

  characterised by the method used for learning · CPC title

• F02D41/0077

  Control of the EGR valve or actuator, e.g. duty cycle, closed loop control of position (EGR valve position sensor F02M26/48) · CPC title