This thesis presents the development of a modular real-time co-simulation framework for vehicle dynamics and autonomous driving applications. The proposed architecture integrates three modules: an autonomous stack, a real-time vehicle simulator core and a photorealistic rendering environment based on the CARLA Unreal engine for sensor emulation and scene management. Communication between layers is achieved through UDP-based interfaces with unified time synchronization, enabling deterministic, closed-loop experiments in both rendered and headless modes. The dynamic model implemented within the simulator describes a 14-degree-of-freedom vehicle, incorporating load transfer, nonlinear combined-slip tire behavior, quasi-steady aerodynamics, and suspension dynamics. Actuator subsystems, namely steering, braking, and powertrain, are modeled as control-oriented systems with delays, saturations, and bandwidth limitations, ensuring identifiability and real-time execution. A dedicated experimental pipeline was defined to identify the key parameters required by the presented simulation model, including Anti-Roll Bar (ARB) stiffness and damping, yaw inertia, and selected tire characteristics. The workflow was implemented within the reference virtual environment (VI-CarRealTime) and designed around measurable vehicle quantities, enabling a direct transfer to real-vehicle testing. The developed framework achieves deterministic timing suitable for Software- and Hardware-in-the-Loop operation and supports realistic perception testing through physics-based sensors. It therefore bridges high-fidelity vehicle dynamics simulation and perception-driven autonomy evaluation within a unified, real-time environment.
La presente tesi presenta lo sviluppo di un ambiente di co-simulazione modulare in tempo reale per applicazioni di dinamica del veicolo e guida autonoma. L’architettura proposta integra tre moduli principali: uno stack autonomo, un simulatore veicolo in tempo reale e un ambiente di rappresentazione fotorealistico basato sul motore CARLA Unreal, utilizzato per l’emulazione dei sensori e la gestione dello scenario. La comunicazione tra i livelli avviene tramite interfacce basate su UDP con sincronizzazione temporale unificata, consentendo esperimenti in anello chiuso a temporizzazione deterministica sia in modalità renderizzata che non. Il modello dinamico implementato nel simulatore descrive un veicolo a 14 gradi di libertà, includendo il trasferimento di carico, il comportamento combinato non lineare degli pneumatici, l’aerodinamica quasi stazionaria e le dinamiche delle sospensioni. I sottosistemi di attuazione (sterzo, frenante e propulsivo) sono modellati come sistemi orientati al controllo, con ritardi, saturazioni e limitazioni di banda, garantendo identificabilità ed esecuzione in tempo reale. È stata inoltre definito un piano di lavoro sperimentale dedicato all’identificazione dei parametri chiave del modello di simulazione, comprendenti la rigidezza e lo smorzamento delle barre antirollio, il momento d’inerzia d’imbardata e specifici parametri pneumatici. Tale sezione è stata sviluppata all’interno dell’ambiente virtuale di riferimento (VI-CarRealTime) e progettata intorno a grandezze misurabili su veicoli reali, in modo da consentirne il trasferimento diretto a test sperimentali su pista. L'ambiente sviluppato garantisce una temporizzazione deterministica, compatibile con configurazioni Software- e Hardware-in-the-Loop, e supporta test di percezione realistici tramite sensori fisicamente modellati. Esso rappresenta un collegamento diretto tra la simulazione ad alta fedeltà della dinamica veicolare e la validazione di sistemi di autonomia basata sulla percezione, all’interno di un ambiente unificato e operativo in tempo reale.
Design and development of a modular co-simulation framework for autonomous vehicles using the CARLA engine
Artese, Gabriele
2024/2025
Abstract
This thesis presents the development of a modular real-time co-simulation framework for vehicle dynamics and autonomous driving applications. The proposed architecture integrates three modules: an autonomous stack, a real-time vehicle simulator core and a photorealistic rendering environment based on the CARLA Unreal engine for sensor emulation and scene management. Communication between layers is achieved through UDP-based interfaces with unified time synchronization, enabling deterministic, closed-loop experiments in both rendered and headless modes. The dynamic model implemented within the simulator describes a 14-degree-of-freedom vehicle, incorporating load transfer, nonlinear combined-slip tire behavior, quasi-steady aerodynamics, and suspension dynamics. Actuator subsystems, namely steering, braking, and powertrain, are modeled as control-oriented systems with delays, saturations, and bandwidth limitations, ensuring identifiability and real-time execution. A dedicated experimental pipeline was defined to identify the key parameters required by the presented simulation model, including Anti-Roll Bar (ARB) stiffness and damping, yaw inertia, and selected tire characteristics. The workflow was implemented within the reference virtual environment (VI-CarRealTime) and designed around measurable vehicle quantities, enabling a direct transfer to real-vehicle testing. The developed framework achieves deterministic timing suitable for Software- and Hardware-in-the-Loop operation and supports realistic perception testing through physics-based sensors. It therefore bridges high-fidelity vehicle dynamics simulation and perception-driven autonomy evaluation within a unified, real-time environment.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_12_Artese_ExecutiveSummary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
10.36 MB
Formato
Adobe PDF
|
10.36 MB | Adobe PDF | Visualizza/Apri |
|
2025_12_Artese_Tesi.pdf
non accessibile
Descrizione: Thesis
Dimensione
50.65 MB
Formato
Adobe PDF
|
50.65 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/247142