On July 4^th, Dr. Alexander L. Gratzer was awarded the 3^rd prize of the Vehicle Association Anniversary Foundation 2024 by the Austrian Association of the Automotive Industry for his dissertation on “Safe and Efficient Model-Predictive Vehicle Control in Mixed and Automated Traffic, öffnet eine externe URL in einem neuen Fenster” written at the Institute of Mechanics and Mechatronics, TU Wien.

As part of the FFG-funded research projects ConnAT, öffnet eine externe URL in einem neuen Fenster and IntIntSec, öffnet eine externe URL in einem neuen Fenster, Alexander’s work presents advanced control strategies for connected and automated vehicles (CAVs) to navigate complex traffic scenarios safely and efficiently.

Model-Predictive Platoon Control

To tackle the challenge of maintaining tightly packed and string-stable heavy-duty vehicle platoons, Alexander proposes a safety-enhanced distributed model-predictive control (DMPC) framework. This concept incorporates a novel time gap spacing policy and optional vehicle-to-vehicle (V2V) communication to ensure robust, collision-free, and string-stable semi-automated truck platooning. String stability is a critical system property of vehicle platoons, describing how disturbances, such as braking or acceleration maneuvers of the lead vehicle, affect the following vehicles. The phenomenon of phantom traffic jams on highways is a result of non-string stable traffic system behavior, typically caused by insufficient inter-vehicle spacing and human driving errors.

Optimal Obstacle Avoidance and Lane Selection

Expanding to 2D trajectory planning and motion control for CAVs, Alexander enhanced his control concepts with implicit obstacle avoidance and lane selection capabilities. To address the challenges of real-time, globally optimal model predictive obstacle avoidance in complex traffic, he formulated a mixed-integer quadratic programming (MIQP) MPC for global optimality and combined it with a quadratic programming (QP)-based MPC in a two-layer control architecture. The QP-MPC ensures local obstacle avoidance, asynchronously updated by the MIQP-MPC solution. Both MPCs use V2V communication for position predictions, while the MIQP-MPC is augmented with lane selection capabilities for optimal lane changes and overtaking. Additionally, he introduced a method to reduce computational complexity for real-time execution.

Dedicated Simulation Framework

For the efficient testing and validation of these advanced control concepts, Alexander created a dedicated multi-agent simulation framework in MATLAB that enables the simulation of complex and highly dynamic traffic scenarios together with the capability to perform co-simulation studies with high-fidelity driving simulators, e.g. IPG Truckmaker or CARLA Simulator.

More Information

• Safe and Efficient Model-Predictive Vehicle Control in Mixed and Automated Traffic
• ConnAT
• IntIntSec
• WKO Österreich, Fachverband Fahrzeugindustrie