Abstracts Track 2026


Area 1 - Intelligent Transport Systems and Infrastructure

Nr: 46
Title:

A Human-in-the-Loop Digital-Twin Investigation of Bicycle-Vehicle Conflicts under Mixed Connectivity

Authors:

Aleksandar Stevanovic, Marko Vukojevic, Ismet Goksad Erdagi, Milan Knezevic, Zeyu Mu and B. Brian Park

Abstract: Although originally developed for vehicle-centric applications, connectivity technologies show promise for improving awareness of nearby road users and mitigating potential conflicts. Their scope now extends to vulnerable road users (VRUs), including bicyclists. These technologies involve various communication paradigms between vehicles, bicyclists, and infrastructure, increasingly designed to address safety challenges for both drivers and cyclists. However, full connectivity is unlikely in the near term, making it essential to study mixed connectivity environments, networks where some but not all users are connected. Such studies should examine how road users experience situations in different connectivity roles, both connected and unconnected, using objective and subjective performance measures. Driving simulators, and more recently, bicycle simulators, have been extensively utilized to: 1. evaluate engineering interventions aimed at enhancing traffic performance, and 2. analyze road user behavior. Prior research in this domain has predominantly focused on simulations involving a single human agent, wherein interactions occur with virtual agents governed by the simulation platform. To complement the previous research, this study presents a novel approach where a unique simulator framework is used, in which two human agents, a driver and a bicyclist, interact within a simulation. Specifically, the study investigates potential conflicting events between a human-controlled driver and a human-controlled bicyclist under various levels of connectivity. Moreover, while most prior driving and bicycle simulator studies have evaluated the safety benefits of connected technologies using hypothetical networks, this study develops a high-fidelity Digital Twin. To replicate an authentic multimodal traffic environment, the study uses a heavy multimodal Delaware Avenue, a two-lane one-way eastbound roadway, located in downtown Newark, DE, USA. A Digital Twin is powered by two synchronized co-simulation platforms: 1. CARLA (Car Learning to Act), an open-source simulator, is employed to provide an immersive and realistic experience for human participants, and 2. PTV Vissim 2025 manages the surrounding background traffic environment. This research firstly addresses a fundamental question: can an experimental setup that integrates a high-fidelity Digital Twin with a dual human-in-the-loop simulator in Virtual Reality successfully reproduce severe conflicts between a bicyclist and a driver? The evaluation was conducted in two stages. First, the study assessed whether the simulated bicyclist–vehicle interactions met established surrogate safety thresholds for severe conflicts. Second, it examined whether participants perceived these events as conflicts through analysis of their self-reported feedback and physiological responses. The subsequent stage of the analysis investigated bicyclist–vehicle conflicts within a mixed connectivity environment, focusing on how collision warning information influenced physiological stress responses, perception and anticipation abilities, and how connectivity affected users’ decision-making.

Area 2 - Intelligent Vehicle Technologies

Nr: 21
Title:

Centralized Nonlinear Model Predictive Control for Coupled Thermal Management Systems in Electric Vehicles

Authors:

Marcell Miszneder and Ulrich Nieken

Abstract: Efficient thermal management and heating, ventilation, and air-conditioning (HVAC) control in hybrid and battery electric vehicles are essential for energy efficiency, operational safety, and battery lifetime. Modern electric vehicles comprise multiple thermally coupled subsystems operating at different temperature levels, such as electric motor and high-voltage battery cooling circuits and the cabin HVAC system. These subsystems are typically controlled in a decentralized manner, which neglects strong thermal interactions and shared components such as the chiller. As a result, unnecessary heat transfer between subsystems may occur, leading to reduced overall efficiency and potentially adverse safety effects, particularly for the high-voltage battery. This work presents an integrated, centralized control concept for coordinating multiple thermally coupled subsystems with different temperature levels by explicitly accounting for their mutual interactions through the chiller. A unified nonlinear model predictive control (NMPC) strategy is proposed for the coordinated control of a coolant-based thermal management system (TMS) and a refrigerant-based HVAC system. The controller simultaneously regulates electric motor, high-voltage battery, and cabin temperatures, while preventing undesired heat transfer from the electric motor to the high-voltage battery, which could compromise battery safety and lifetime. The scope of this study is limited to cooling operation; heating modes and thermal management under cold ambient conditions are not considered. To enable accurate and numerically stable prediction within the NMPC framework, a high-fidelity yet computationally efficient thermal model is developed. The model is based on reduced-order first-principles equations and captures the thermal inertia of different components and working fluids, as well as their interaction within the chiller. The phase-change effects are approximated using a smoothed, transfer-function-based formulation, resulting in differentiable system dynamics suitable for gradient-based NMPC optimization. The NMPC control objectives include accurate and stable temperature tracking of all subsystems, energy-efficient operation, and smooth actuator behaviour under operational and physical constraints. By efficiently combining and coupling the coolant loops, the proposed strategy minimizes unnecessary heat exchange between subsystems and improves the overall thermal efficiency of the vehicle. A comprehensive parametric study is conducted to systematically determine suitable prediction and control horizons and weighting factors in the NMPC cost function. A structured parametrization approach was developed, which enables faster convergence to effective parameter sets while maintaining balanced performance with respect to temperature regulation, energy consumption, and actuator usage. Closed-loop simulations under standardized driving cycles and varying ambient conditions demonstrate the effectiveness of the proposed single-controller, multi-subsystem NMPC framework. The results show robust temperature regulation, improved energy efficiency, and enhanced safety-relevant thermal behaviour, highlighting the potential of centralized predictive control for advanced thermal management in future electric vehicles.