Smart vehicle control device and method

A smart vehicle control method performed by a computing device including a processor and a memory, the method comprising:

• receiving, by the processor, a user input from a user terminal through a digital key application;
• receiving, by the processor, a vehicle input from a vehicle by a user boarding on the vehicle;
• performing, by the processor, reinforcement learning based on the user input and the vehicle input by use of a reinforcement learning model, and generating, by the smart vehicle control device, a vehicle control policy for vehicle driving preparation;
• and
• controlling, by the processor, the vehicle based on the vehicle control policy,
• wherein the performing, by the processor, the reinforcement learning based on the user input and the vehicle input by use of the reinforcement learning model, and generating, by the smart vehicle control device, a vehicle control policy for vehicle driving preparation comprising:
• upon receiving a user answer to a question related to the vehicle driving preparation provided to the user through the digital key application, updating an initial vehicle control policy by performing the reinforcement learning model based on a first state and a first reward, generated based on the received answer,
• upon receiving a real-time user input related to the vehicle driving preparation through the digital key application, updating an initial vehicle control policy by performing the reinforcement learning based on a second state and a second reward, generated based on the real-time user input,
• upon receiving the vehicle input which is input from the vehicle by the user boarding in the vehicle, updating the initial vehicle control policy by performing the reinforcement learning based on a third state and a third reward, generated based on the vehicle input, and
• generating the vehicle control policy by repeatedly updating the initial vehicle control policy for reward including the first, second, and third rewards to include a maximum value,
• wherein the first reward includes a value less than zero, the third reward is a minimum negative value, the second reward is a median negative value, and the first reward is a maximum negative value.