Agent-based development of natural transportation networks.

摘要

This thesis presents a simulation for generating transportation networks. The generated transportation networks have a natural character developed through successive agent trips, where the individual plan of each agent is incorporated into the network. The agents are like social insects, which communicate by leaving signs in the environment and collectively build structures. Network links in the simulation are gradually moved to locations which benefit most agents, generating an efficient and robust network. This thesis presents a dynamically-modified transportation network as a data structure and set of operations. The technique is applied to model leaf venation and urban transportation networks. Through the software interface, a user can control the parameters and a three-dimensional visualization. By exploring how agent parameters affect the character of the generated networks, the applicability of nature-inspired algorithms in urban network planning is shown.