Abstract

In the last decades, vehicles have evolved from mechanical machines to sophisticated hardware and software (HW/SW) dominated systems. This digital paradigm shift can be explained with the rise of Advanced Driver Assistance Systems (ADAS), as well as applications of highly automated driving. However, such features require a nearly exponential amount of additional SW functionalities. To support the performance demands of such applications, powerful multi and many-processor HW platforms have been employed, exhibiting a trend toward more integrated architectures. Yet, as a consequence, the design and validation, as well as safety assessment of the digital in-vehicle infrastructure of modern cars has become immensely complex.Several techniques arose to bridge this gap, e.g., system engineering strategies, and model-based design and validation methods. In this thesis, it is proposed to augment such traditional approaches with purely simulation-driven techniques to improve and accelerate overall system design. Firstly, the utilization of driving simulators is suggested to conduct virtual road tests in an early application design stage. Furthermore, virtual prototyping is proposed to perform system exploration and extend the simulation scope of ADAS to the embedded HW/SW implementation.However, the main research contributions of the thesis are two techniques that enable efficient full-vehicle modeling. Firstly, multi-scale simulation, an approach to integrate submodels of a top-level simulation system on different design abstractions. Its main benefit is providing a tunable resolution for models, thus enabling a trade-off between the execution performance and the required simulation accuracy. Secondly, multi-domain co-simulation is proposed, a standardized method to interconnect and jointly utilize multiple vehicular simulation ecosystems. This approach addresses the multi-disciplinary nature of vehicles, requiring specialized simulators that are best suited to implement a formalism representing an individual domain.Putting all aforementioned concepts in practice, a comprehensive fully virtual ADAS prototyping ecosystem is presented in this thesis. The proposed framework assembly covers various phases of the ADAS development flow, such as model-based design, virtual system integration, virtual system testing and verification, as well as application refinement. To highlight its potentials, a multitude of ADAS applications are rapidly prototyped via the proposed system. Lastly, insights into application and system-level analyses are presented, while simulation performance evaluation is provided to highlight the acceleration potential of the virtual ecosystems.

Abstract

Bonded-particle modelling is a powerful approach for simulating particle made structures with enhanced abilities, i.e., structures that are able to deform, brake/fracture, and thus interact with other particles of a DEM simulation in an enhanced complex manner. This modelling approach is based on connecting particles with beam elements (commonly termed bondings, with each bonding connecting two particles) to create a bonded-particle model (BPM), typically consisting of a large number of particles and bondings. The behaviour of such a BPM can be adjusted in two categories: via properties related to the particles, e.g. their mass, shape, or frictional characteristics, and via properties related to the bondings, e.g. their Young’s moduli or breakage criteria. This contribution presents developments in complex bonded-particle modelling for achieving flexible/deformable and breakable/fracturable structures and highlights the enhanced capabilities made possible by using this approach. For this purpose, various applications for bonded-particle modelling in DEM simulations are presented, with selected case studies each demonstrating the effectiveness of bonded-particle modelling in solving practical engineering problems, including: - 1D BPMs, e.g., deformable beams or ropes and chains, - 2D BPMs, e.g., membranes, textiles, nets, bags, shell-like parts (e.g. silos), or conveyor belts (with particular reference to dynamic belt simulation), and - 3D BPMs, e.g., complex-breakable structures (e.g. with considering crack formation or dynamic impact handling; e.g. illustrated on filter cake material). Furthermore, possible interaction scenarios, including interactions of BPMs to particles, rigid parts, other BPMs, or even in terms of SPH (Smoothed-Particle Hydrodynamic) or MBD (Multibody Dynamics), are addressed. Finally, current developments and potential future directions in this field are outlined, and potentials for extending this modelling approach to further application areas are discussed.

Abstract

EuCAP 2016 - 10th European Conference on Antennas and Propagation, Davos, SUISSE, 10-/04/2016 - 15/04/2016; With Advanced Driver Assistance Systems (ADAS) getting always more sophisticated, the related Virtual Prototyping platforms have to propose a very high level of accuracy with improved flexibility regarding vehicles, sensors, environments and scenarios. In this paper a new strategy is introduced for RADAR sensors modeling aimed at allowing high accuracy while limiting the related development efforts.