Multi-agent Spatiotemporal Simulation of Autonomous Vehicle Fleet Operation

Abstract

Autonomous vehicle fleets, consisting of self-driving vehicles, are at the forefront of transportation innovation. The appearance of autonomous vehicles (AVs) provides a new solution for traffic problems and a new market for transportation network companies such as DiDi and Uber. Conducting simulations in the present is indeed crucial to prepare for the eventual operation of autonomous vehicles, as their widespread adoption is expected to occur in the near future. This research adopts an Agent-Based Modelling (ABM) approach to understand and optimize the performance of autonomous vehicle systems. Moreover, Geographic Information System (GIS) technology also plays a crucial role in enhancing the effectiveness and accuracy of the simulation process. GIS enables the representation and manipulation of geospatial data, such as road networks, land-use patterns, and population distribution. The combination of ABM and GIS allows for the incorporation of real-world geographic data, providing a realistic and geographically accurate environment for the agents in the virtual environment. In this thesis, the multi-agent spatiotemporal simulation is conducted by the GAMA platform. The model simulates the behaviour and interactions of individual agents, which are fleet agents and commuters, to observe the emergent behaviour of the entire system. Within the experiment, different scenarios are considered for both people and fleets to explore a range of approaches and strategies. These scenarios aim to evaluate the effectiveness of various approaches in meeting dynamic commute needs and optimizing fleet operations. By simulating these different scenarios and analyzing their outcomes, the study aims to provide insights into the improvement of fleet size and deployment in autonomous vehicle systems. The ultimate goal is to identify effective strategies that lead to optimized fleet size in different scenarios, reduced idling time and emission, improved traffic management, and overall more efficient and sustainable autonomous vehicle systems.