Can an AI agent judge when you talk to you while driving? According to the KAIST research team, the in-vehicle conservation service technology will judge when it is appropriate to contact you to ensure your safety.
Professor Urchin Lee of the Department of Industrial and Systems Engineering at KAIST and his research team have developed AI technology that automatically detects safe moments for AI agents to provide conversational services to drivers.
His research focuses on solving potential problems of distraction created by in-vehicle conversation services. If an AI agent talks to a driver at an inconsistent time, such as when making a turn, a car accident is more likely to occur.
Conversation services in the vehicle need to be secured as well as secure. However, the cognitive burden of multitasking negatively affects service quality. Users are more distracted during certain traffic situations.
To address this long-standing challenge of in-vehicle interaction services, the team introduced a holistic cognitive model, which considers both safe driving and auditory-verbal service performance, and a machine-learning model for all collected data uses it.
The combination of these individual measures is capable of determining the appropriate moments for negotiation and the most appropriate type of conversational services.
For example, in the case of providing simple-reference information such as meteorological forecasts, driver safety alone would be the most appropriate consideration.
Meanwhile, a combination of driver safety and auditory-verbal performance should be considered when a driver’s response is required, such as “yes” or “no”.
The research team developed a prototype of the in-vehicle conversation service based on a navigation app, which can be used in real driving environments.
The application was also connected to the vehicle to collect OBD-II / CAN data in the vehicle, such as steering wheel angle and brake pedal position, and mobility and environmental data such as the distance between consecutive cars and traffic flow.
Using pseudo-negotiation services, the research team collected real-world driving datasets consisting of 1,388 interactions and sensor data from 29 drivers who interacted with AI conversational agents.
Machine learning analysis based on the dataset demonstrated that the appropriate moment for driver interruption can be correctly estimated with 87% accuracy.
Safety enhancement technology developed by the team is expected to reduce driver distractions caused by in-vehicle negotiation services. This technology can be applied directly to current vehicle systems that provide negotiation services.
This can be extended and applied to detect driver distraction problems caused by the use of a smartphone while driving.
Professor Lee said, “In the near future, cars will provide conversational services in various types of vehicles.
This technology will definitely help vehicles to safely negotiate with their drivers, as it will only be using fixed sensor data Negotiation can determine the services to be provided fairly accurately. Cars. “