| dc.description.abstract | The goal of this thesis is to develop a system that improves urban transit networks. A good transit network can reduce the number of vehicles on the road as people will favor public transport over private transportation. This will eventually help reduce congestion and environmental emissions.
The Urban Transit Routing Problem (UTRP) concerns the creation of route networks. UTRP is a complex and multiconstrained problem, in which the creation of route networks can both be challenging and time consuming. Metaheuristics like swarm intelligence methods have proven to be effective of finding sufficient solutions to these types of NP-hard problems. In this contribution, a swarm inspired optimization system is designed and presented, aiming to create efficient solutions to the UTRP. The proposed system uses an ant colony approach with, unlike previous techniques, additional attributes inspired by bee colony optimization and particle swarm optimization.
A structured literature review is conducted to synthesize the relevant primary studies. All retrieved results are presented and analyzed. Further, because metaheuristics require good parameter values to solve concrete problems optimally, a thorough review and justification of each selected parameter is documented. This documentation will contribute in providing a starting point for potential future research. A comparison of a standard ant colony optimization (ACO) implementation is performed to determine whether the proposed system improves the standard ACO. To demonstrate the performance of the proposed system, obtained results are compared against results published in the literature. Results are compared on the basis of Mandl's benchmark problem, which is a widely investigated and acknowledged benchmark problem. The proposed system is also tested on larger networks, more similar to real transit networks, to validate whether the proposed system supports larger networks as input. This thesis will also report how the usage of the graph database Neo4j has affected the development and quality of the proposed system.
Comparison of obtained results with the standard ACO implementation and other published results are promising, especially regarding the average traveling time per transit user. | |
