Vehicle Dynamic Model for Real-Time Applications and Training of Artificial Drivers

This thesis presents the development of a full vehicle model tailored for real-time simulation applications. A comprehensive modular simulation framework is developed with the primary goal of providing an accurate and flexible vehicle model for real-time simulations. The vehicle model is developed using a multibody dynamics approach, leveraging efficient formulations and symbolic manipulation to generate cost-effective analytical models. This work presents the model's underlying theory and practical implementation, showcasing the framework's modularity that facilitates seamless integration of external models. This adaptability enables the model's application in a wide range of scenarios, ranging from vehicle dynamics analysis to the development of advanced driver assistance systems. The simulation framework is utilized for Hardware-in-the-Loop (HIL) and Driver-in-the-Loop (DIL) simulations, proving its efficacy in real-world scenarios. Validation against commercial simulation software and real-world vehicle telemetry data further corroborated the model's fidelity and accuracy. Beyond the core vehicle model, this research introduces innovative models that significantly contribute to the simulation efficiency and accuracy. A novel Limited Slip Differential (LSD) model is proposed, employing a smooth equation switching approach to improve numerical efficiency and address issues of conventional approaches such as robustness and the non-exact representation of the locked state. Furthermore, a Symbolic-Numerical Approach is presented for analyzing structure compliance, with a specific application to vehicle suspension compliance analysis. This approach exploits the symbolic manipulation capabilities of computer algebra software to generate efficient analytical models for the suspension's Kinematics and Compliance (K&C) characteristics. This approach not only provides a comprehensive methodology for analyzing suspension compliance but also offers a framework for the generation of symbolic models for any compliant mechanisms, enhancing efficiency in simulation, design, sensitivity analyses, and optimization tasks. In essence, this Ph.D. thesis presents a comprehensive vehicle model for real-time simulation applications, featuring a flexible and modular framework alongside novel models that address specific challenges in vehicle dynamics simulation. Real-time capabilities of the developed model enable closed-loop simulations making it a powerful tool for virtual prototyping, performance evaluation, and controller development in the automotive domain. Practical applications in the field of autonomous vehicles and advanced driver assistance systems showcase the applicability of the proposed framework and models, offering a user-friendly framework for future research and development.