Brief summary: Within this thesis, different modeling approaches at different scales in the domains of urban radio propagation, decentralized channel coordination, and information dissemination in inter-vehicle communication networks are investigated. The contributions reveal the suitability of existing models for network-oriented research, propose a novel information-centric modeling approach, and identify characteristics of inter-vehicle communication systems which determine key dependability aspects.
Inter-vehicle communication is considered a promising approach to increase road safety and efficiency in the near future. Vehicles which do not solely have to rely on their own sensors to assist the driver, but are able to incorporate the communicated perception of other road users are able to make more profound decisions or are even able to negotiate cooperative driving strategies in a subsequent step.
Given the intended use cases of inter-vehicle communication as well as the harsh environment it is going to be deployed in, dependability of the communication system is of central importance for the development and decisive for the assessment of the overall system. Simulation-based assessment has hereby made important contributions, by evaluating and comparing protocols, identifying bottlenecks, or performing impact assessment. For this purpose, a variety of different simulation models have been proposed in the past, some of which address the same aspect, but on a different abstraction level or with some of them building on different assumptions. However, while on the one hand it is not always clear which situations are appropriately represented by which model and whether multiscale consistency is ensured across multiple layers, on the other hand, there is yet a shortcoming of mesoscopic and macroscopic models for the simulation of inter-vehicle communication. As a consequence, i) there is still uncertainty about the large-scale benefits of Vehicular Ad-hoc networks, and ii) a deeper understanding is missing in which situations the envisioned inter-vehicle communication system has its shortcomings with respect to dependability.
The contributions of this thesis are motivated by the objective of providing dependable inter-vehicle communication systems by demonstrating and contrasting different modeling approaches as well as increasing the understanding of how particular conditions or protocol features determine the characteristic and performance of inter-vehicle communication networks. Given the wide field of activities, we narrowed down our consideration to the following three multiscale modeling aspects, which decisively influence the dependability of the overall system:
- From radio propagation to reception capability: Urban radio propagation conditions decisively influence how far and reliable vehicles can communicate in an urban environment and hence determine the fields of application of the envisioned inter-vehicle communication system. We first give a survey over urban radio propagation models which are currently employed within the vehicular networking community and discuss their applicability and validity. The principles of ray tracing, a fine-grained modeling approach, are discussed and we present how it can be applied for the purpose of dependability assessment in inter-vehicle communication research. Subsequently, we employ ray tracing on the one hand to evaluate which simplified model yields appropriate results in which situations and on the other hand to study the general sensitivity of urban radio propagation conditions to various obstructions and hence to identify challenging conditions for future urban inter-vehicle communication applications. Our results show that even though situations, such as vehicles obstructing the line-of-sight, cannot be explicitly modeled by most currently employed statistical models, average radio conditions are appropriately represented, even in densely populated urban scenarios. The achieved results enable to assess the dependability of urban inter-vehicle communication by not just considering mean values, but also occurring outliers in future studies.
- From medium access control to packet reception: The dependability of packet exchange in inter-vehicle communication networks heavily depends on the ability of the employed coordination scheme to successfully coordinate the channel access between participating stations under a variety of occurring circumstances, such as occurring decoding errors or the existence of interfering stations. For this reason, we assess and contrast the dependability of two prominent but fundamentally different decentralized coordination schemes -- IEEE 802.11p Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) and Self-organizing Time Division Multiple Access (STDMA). In order to facilitate a profound analysis of STDMA we contribute a comprehensive protocol description and discussion of existing turning knobs as well as protocol elements that govern its function. We pursue a fundamental analysis of the STDMA protocol by first identifying situations in which, even though two stations are within each other's communication range, the STDMA protocol might not be able to successfully coordinate the channel, i.e., prevent concurrent transmissions by multiple stations. By means of an extensive simulation study we then quantify the likelihood of these situations to occur under a variety of parameters. These results help us to identify i) the main characteristics of STDMA in dealing with various situations and ii) how STDMA's protocol parametrization influences its capability to successfully coordinate the channel access in a vehicular environment. Since fundamental differences in strengths and weaknesses of both medium access control protocols became apparent, we subsequently investigate their cause and impact on a network level in more detail. In particular, we focus on understanding to which degree communication conditions, such as the severity of fading conditions, and mobility influences the coordination ability of STDMA compared to CSMA/CA and we hence contribute to an in-depth assessment of its dependability. Our results show that on the one hand the reservation-based approach of STDMA is beneficial under fluctuating channel conditions, since it allows to incorporate reservation information received in the past, on the other hand it is detrimental whenever there is a change of the vicinity, since STDMA requires more time to adjust to altered circumstances. Despite these differences, the obtained results justify to abstract from a specific protocol realization for macroscopic modeling approaches.
- From packet level to information level: The availability of information determines whether a vehicle is reliably able to give advices to its driver. In order to address this issue, we propose an information-centric modeling concept which allows to directly model the availability of information in an inter-vehicle communication network without taking the detour of first having to model each and every packet exchanged in the network, as is done nowadays. Motivated by the vision to enable large-scale assessments, we subsequently present a light-weight modeling approach which can be employed within this modeling concept and allows to describe the dissemination delay within arbitrary inter-vehicle communication networks. The approach is based on the analogy of a construction kit where the road network is divided into elementary road segments, such as straight roads or intersections and whose characteristics w.r.t. the time it takes an information to traverse these blocks is known in advance. Determining the information delay is then just a matter of determining the shortest path between the information source and destination. In order to characterize and determine the traverse time for each block we perform an extensive simulation study, incorporating parameters related to the road network layout, mobility of the vehicles, communication conditions, and the data dissemination protocol. By means of an evaluation scenario, consisting of several of these building blocks, we show the general feasibility of this approach and discuss its applicability to enable future large-scale impact assessments of inter-vehicle communication.