City Research Online

Abstract

Fully automated vehicles are expected to have a significant share of the road network traffic in the near future. Several commercial vehicles with full-range adaptive cruise control systems or semi-autonomous functionalities are already available in the market. This provides a unique opportunity to improve the acceleration behaviour of vehicles, and thereby, improve network’s efficiency in terms of important performance indicators such as fuel consumption and traffic throughput. However, automated driving systems usually adopt a highly conservative driving strategy to ensure safety and fuel efficiency for individual vehicles. The collective impacts of such strategies on the network level can lead to the deterioration of traffic flow and to an increase in fuel/energy consumption. Much of the existing research in this area either target driving conditions where there are no additional complexities caused by interaction between vehicles, or make simplistic assumptions about the dynamics of driving behaviour and its relationship with fuel consumption in order to formulate a feasibly solvable optimisation problem.

The reduction of the question of fuel efficiency to optimisation scenarios where only a pair of vehicles are considered and little attention is paid to the surrounding traffic, leads to a user-optimal driving strategy at best, however addressing environmental concerns and a more efficient use of fossil fuels in road transport networks necessitates a system-optimal approach. A system-optimal approach means the scope of the problem must be broadened so that a) the complex relationship between individual driving styles and the dynamical features of traffic flow are incorporated within the optimisation framework and b) the long-term impacts of driving strategies on network’s performance are modelled within optimisation scenarios. The challenge here is to model driving behaviour and traffic flow with sufficient accuracy and devise an optimisation framework that is computationally efficient enough to cope with the complexity of the problem.

In this study, the use of car-following models and limiting the search space for optimal strategies to the parameter space of car-following models is proposed. This framework enables performing much more comprehensive optimisations and conducting more extensive tests on the collective impacts of fuel-economy driving strategies. The results obtained in this study show that formulating the optimisation in a short-sighted way where merely individual vehicles are considered and no attention is paid to the collective impacts of a fuel-economy driving strategy, can lead to significant increase in fuel consumption for the whole network while delivering marginal benefits for the target vehicle. This study establishes a complex relationship between traffic flow and fuel consumption on the link level, where the former cannot be achieved without addressing the latter correctly in the optimisation process.

In addition to the main research question discussed above, the present thesis proposes a new method for the analysis of car-following models. The conventional method in the analysis of car-following models relies on the cumulative error between real data and modelled data in order to benchmark car-following models. Although the cumulative error is indeed an informative measure of performance, it leaves many questions regarding the capacity of models for replicating driving behaviour unanswered. Here the use of dynamic system identification is investigated as a way to provide a more in-depth analysis of the strengths and weaknesses of car-following models in reproducing realistic driving behaviour. Subsequently, the proposed method is applied to compare a number of car-following models. Although the application of this method for the comparison of car-following models presents some challenges that require further research, the results are very promising.

Publication Type:	Thesis (Doctoral)
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TL Motor vehicles. Aeronautics. Astronautics
Departments:	Doctoral Theses School of Science & Technology > Engineering School of Science & Technology > School of Science & Technology Doctoral Theses

Export

Downloads