Development of a Connected and Autonomous Vehicle Modelling Framework, with Implementation in Evaluating Transport Network Impacts and Safety

Abstract

Transportation systems form critical links that connect developed cities with the broader world. They connect our residential, recreational, employment, and natural environments. Annual population and vehicle ownership growth place an increasing strain on transport systems, resulting in escalating levels of congestion and delay. Using new infrastructure and expansion is a problematic solution, as it often incentivises greater private vehicle use and worsens long-term congestion. Infrastructure expansion requires repeated and increasing levels of capital investment.

The Connected and Autonomous Vehicle (CAV) is an emerging technology that facilitates communication with infrastructure and other agents. CAVs address many inefficiencies of human driving by exhibiting instantaneous reaction times, smaller headways, and vehicle platooning. Their fundamentally different driving behaviour may render many infrastructure planning and modelling tools not applicable to future mixed fleets and CAVs. This thesis develops a comprehensive modelling framework for the emulation of CAV behaviour in microsimulation, with a focus on car-following, lane-changing, gap-acceptance, autonomous merging, and vehicle cooperation. The developed framework is implemented in a range of investigations aimed at better understanding the impact of mixed fleets and CAVs on vehicle kinematics, intersection performance, and safety. Uncertainties regarding CAV behaviour and motorway capacity, delay redistribution through signal optimisation, and the need for recalibrating macrosimulation modelling parameters are also investigated.

These investigations indicate that CAVs improve network performance, driver aggression (acceleration), and driver comfort (jerk). Low levels of penetration improved fleet operations, leading to increased throughput, increased capacity, reduced delay, and reduced likelihoods of accidents and conflicts. Average network delay is decreased significantly by redistributing the CAV travel time savings to all network agents, through signalling re-optimisation. Finally, this thesis demonstrates that macrosimulation modelling parameters used for human fleets show reduced predictive qualities when applied to mixed fleets and CAV environments. The performed recalibration provides significantly improved results in the predictive quality of volume-delay functions.