Towards Safe and Reliable Autonomous Driving: A Tutorial on Sampling-Based MPC, Situation Awareness, and Decision-Making Under Uncertainty
November 4 (Tuesday), 13:00-17:30, Premier Ballroom, 2F
Registration Fee: USD 110 / KRW 150,000
Apply for Tutorial: https://sigongji.2025.iccas.org/program/registrant/registrant.asp
*My page- Tutorial Tab
 |
 |
 |
| Prof. Heejin Ahn (KAIST, Korea) Assistant Professor Topics of recent interest include |
Prof. Kyoungseok Han (Hanyang University, Korea) Associate Professor Topics of recent interest include |
Prof. Changmook Kang (Hanyang University, Korea) Associate Professor Topics of recent interest include |
[13:00-14:20] Title: Sampling-based Model Predictive Control
Prof. Heejin Ahn (KAIST, Korea)
This tutorial will introduce sampling-based model predictive control (MPC) methods, focusing on the Model Predictive Path Integral (MPPI) and the Cross-Entropy Method (CEM). By framing the optimal control problem as a Bayesian inference problem, we will derive these algorithms from first principles and highlight their practical strengths and limitations. In addition, we will discuss methods for guaranteeing safety to address the inherent risks of sampling-based approaches. The tutorial will combine theoretical background with simulations and open-source code demonstrations to provide both intuition and practical guidance for applying these methods in research and real-world systems.
[14:35-15:55] Title: Data-Driven Decision-Making and Control of Autonomous Vehicles Under Uncertainty
Prof. Kyoungseok Han (Hanyang University, Korea)
This talk presents recent advances in data-driven control and adaptive decision-making strategies for autonomous vehicles operating in uncertain driving environments. We first highlight how Gaussian Process (GP) regression can be leveraged to capture complex vehicle dynamics and unmodeled effects, thereby improving the fidelity of prediction models in the Model Predictive Control (MPC) framework to improve the trajectory tracking performance. Building on this foundation, we introduce game-theoretic decision-making methods that enable autonomous vehicles to anticipate and respond to the diverse behaviors of heterogeneous traffic participants, facilitating safe and socially aware interaction. Throughout the tutorial, fundamental theories are paired with illustrative examples and MATLAB demonstrations, allowing participants to gain hands-on experience in implementing GP-based modeling, MPC design, and game-theoretic control.
[16:10-17:30] Title: Enhancing Autonomous Mobility Safety via Model-Driven Situation Awareness
Prof. Changmook Kang (Hanyang University, Korea)
This work explores a model-driven approach to strengthen safety and reliability in autonomous mobility. By interpreting diverse driving conditions through multiple vehicle models, the framework enhances understanding of system behavior and environmental context. Field validations show its potential to support safer decision-making and resilient operation across both structured roads and challenging off-road terrains.
