Control device, control method, and recording medium

1. A control device, comprising:

a memory; and

a processor coupled to the memory, the processor configured to:

obtain a driving condition and an actual value of a controlled amount of an engine under control,

calculate an estimate value of a future controlled amount, from the obtained driving condition and a candidate value of an operation amount for controlling the controlled amount, the processor using a transformed engine model to calculate the estimate value, the transformed engine model in which an activation function of each neuron is transformed using a linear inequality function having a binary variable, the transformed engine model for which a weight coefficient, bias, and upper and lower limits of an input/output variable of the each neuron are set based on an engine model that has a multilayer neural structure and the activation function of a ReLU structure, receives the driving condition and the operation amount of the engine as input, and predicts dynamic characteristics of the engine,

determine a value of the operation amount for controlling the controlled amount, based on an error between the calculated estimate value of the controlled amount and a target value of the controlled amount identified from the driving condition, and

output the determined value of the operation amount.

2. The control device according to claim 1 , wherein

the processor determines the value of the operation amount, based on the error between the target value of the controlled amount and the estimate value of the controlled amount, and a difference of a current value of the operation amount and the candidate value of the operation amount.

3. The control device according to claim 2 , wherein

the candidate value includes a plurality of candidate values, and

the processor:

based on the error and the difference for each of the plurality of candidate values, calculates, for each of the plurality of candidate values, a control evaluation value that represents a cost for controlling the controlled amount, and

based on the control evaluation value calculated for each of the plurality of candidate values, determines the value of the operation amount for controlling the controlled amount from among the plurality of candidate values of the operation amount.

4. The control device according to claim 1 , wherein

the controlled amount is at least any one of an intake manifold pressure, an intake air volume, and an intake manifold oxygen concentration.

5. The control device according to claim 4 , wherein

the driving condition includes an engine rotation speed and a fuel injection amount.

6. The control device according to claim 5 , wherein

the operation amount is at least any one of a turbo position, a throttle position, and an exhaust gas recirculation opening.

7. The control device according to claim 1 , wherein

the upper and lower limits of the input/output variable are set based on a maximum value and a minimum value of the input/output variable obtained by providing a chirp signal or an APRBS signal to the engine model as input.

8. The control device according to claim 1 , wherein

the activation function is a Leaky ReLU structure.

9. The control device according to claim 1 , wherein

the transformed engine model is updated based on training data that indicates the actual value of the controlled amount corresponding to a combination of the driving condition and the value of the operation amount.

10. The control device according to claim 1 , wherein

the processor is configured to:

refer to correction information that indicates correspondence between a plurality of correction values of the controlled amount and a plurality of driving conditions of the engine, the processor identifying, among the plurality of correction values, a correction value that corresponds to the obtained driving condition,

add the identified correction value to the calculated estimate value of the controlled amount and thereby, correct the estimate value of the controlled amount, and

determine the value of the operation amount for controlling the controlled amount, based on an error between the target value of the controlled amount and the corrected estimate value of the controlled amount.

11. A computer-implemented control method, comprising:

obtaining a driving condition and an actual value of a controlled amount of an engine under control;

calculating an estimate value of a future controlled amount, from the obtained driving condition and a candidate value of an operation amount for controlling the controlled amount, the estimate value being calculated using a transformed engine model in which an activation function of each neuron is transformed using a linear inequality function having a binary variable, the transformed engine model for which a weight coefficient, bias, and upper and lower limits of an input/output variable of the each neuron are set based on an engine model that has a multilayer neural structure and the activation function of a ReLU structure, receives the driving condition and the operation amount of the engine as input, and predicts dynamic characteristics of the engine;

determining a value of the operation amount for controlling the controlled amount, based on an error between the calculated estimate value of the controlled amount and a target value of the controlled amount identified from the driving condition; and

outputting the determined value of the operation amount.

12. A computer-readable recording medium storing therein a control program executable by one or more computers, the control program comprising:

an instruction for obtaining a driving condition and an actual value of a controlled amount of an engine under control;

an instruction for calculating an estimate value of a future controlled amount, from the obtained driving condition and a candidate value of an operation amount for controlling the controlled amount, the estimate value being calculated using a transformed engine model in which an activation function of each neuron is transformed using a linear inequality function having a binary variable, the transformed engine model for which a weight coefficient, bias, and upper and lower limits of an input/output variable of the each neuron are set based on an engine model that has a multilayer neural structure and the activation function of a ReLU structure, receives the driving condition and the operation amount of the engine as input, and predicts dynamic characteristics of the engine;

an instruction for determining a value of the operation amount for controlling the controlled amount, based on an error between the calculated estimate value of the controlled amount and a target value of the controlled amount identified from the driving condition; and

an instruction for outputting the determined value of the operation amount.