AUTONEWS

New AI-based technology offers real-time electric vehicle state estimation for safer driving

A research team led by Professor Kanghyun Nam from the Department of Robotics and Mechanical Engineering at DGIST has developed a physical AI-based vehicle state estimation technology that accurately estimates the driving state of electric vehicles in real time.

This technology is viewed as a key advancement that can improve the core control performance of electric vehicles and greatly enhance the safety of autonomous vehicles. The work was conducted through international joint research with Shanghai Jiao Tong University in China and the University of Tokyo in Japan.

The work is published in the journal IEEE Transactions on Industrial Electronics.

The sideslip angle, which shows how much an electric vehicle slides sideways during sharp turns or on slippery surfaces, is crucial information for safe driving. Since this value is difficult to measure directly with in-vehicle sensors, automakers depend on complex physical models or indirect estimates. These methods face issues due to low accuracy and limitations across different driving conditions.

To address these issues, Professor Nam's team created a new physical AI-based estimation technology combining AI and physical models. The main idea is that it significantly enhances accuracy by merging a physical model describing vehicle motion with data from sensors measuring lateral tire force and an AI-based regression model (GPR).

To address nonlinear tire behavior and environmental changes that are difficult for physical models to explain, the research team developed a hybrid estimation framework that combines a physical tire model with an AI-based learning model. Specifically, by using an unscented Kalman filter (UKF) observer integrated with Gaussian process regression (GPR), the team secured both the flexibility of data-driven learning and the reliability of the physical model. This combination enables more accurate and faster estimation of the vehicle's slip angle compared to traditional methods.

During actual electric vehicle platform testing, this technology exhibited high accuracy and strong estimation performance across different road surfaces, speeds, and cornering conditions. Accurate vehicle state estimation is essential for driving stability control, autonomous driving safety, and energy efficiency in electric vehicles. This achievement is seen as a major technological breakthrough for future mobility, as it opens new possibilities for AI-based physical vehicle control.

Professor Nam stated, "Through a new approach that combines physical models and AI, we can estimate the driving conditions of electric vehicles with greater precision and reliability. This research will serve as a core foundation for next-generation autonomous driving and electric vehicle technology. We will further develop this technology through joint research with global automakers and expand it into a technology applicable to actual industrial settings."

In 2025, researchers introduced a breakthrough physical AI-based estimation technology designed to enhance electric vehicle (EV) safety by accurately predicting driving states in real-time. 

Physical AI for Precise Vehicle Control...A research team led by Professor Kanghyun Nam (DGIST) developed a hybrid framework that merges traditional physical vehicle motion models with AI-based regression models (Gaussian Process Regression). 

Key Innovation: The system estimates the sideslip angle—the degree to which a vehicle slides sideways during sharp turns or on slippery surfaces.

Why It Matters: This value is difficult to measure with standard sensors; older methods often failed in low-traction conditions like ice.

Performance: Tested on actual EV platforms, the AI achieved high accuracy across diverse road surfaces, speeds, and cornering conditions. 

Broader AI State Estimation Trends in 2025...Beyond vehicle motion, AI is now fully integrated into EV battery and system management to provide real-time updates: 

State of Health (SOH) & Charge (SOC): New hybrid neural networks (e.g., 2D-CNN combined with self-attention) achieve 98.88% accuracy in estimating battery health, reducing prediction errors by over 80% compared to previous standards.

Predictive Maintenance: AI now monitors real-time current flow and voltage to detect component faults before they cause accidents, reducing roadside incidents by up to 24% in some 2025 fleet deployments.

Enhanced Range: By adjusting to individual driving styles and environmental data in real-time, AI energy optimization has improved range by up to 6% without hardware changes. 

Safety Impacts

Autonomous Driving: Accurate state estimation serves as the "brain" for next-generation autonomous systems, identifying pedestrians and road hazards with up to 92% precision in 2025.

Stability Control: Real-time slip angle data allows for faster, more precise stability control interventions, crucial for preventing accidents on slippery roads. 

Provided by Daegu Gyeongbuk Institute of Science and Technology