Evaluating clustering methods within the Artificial Ecosystem Algorithm and their application to bike redistribution...

Area of Science:

• Operations Research
• Computer Science
• Artificial Intelligence

Background:

• The truck redistribution problem in bike share schemes (BSS) is an NP-hard combinatorial optimization challenge.
• Existing exact optimization techniques are computationally infeasible for real-time solutions.
• Efficient heuristic approaches are needed to manage BSS logistics.

Purpose of the Study:

• To propose and evaluate the Artificial Ecosystem Algorithm (AEA) as a heuristic solution for the BSS truck redistribution problem.
• To adapt the AEA for distributed computing and dynamic problem changes.
• To assess the effectiveness of AEA variants in optimizing BSS operations.

Main Methods:

• Developed three Artificial Ecosystem Algorithm (AEA) variants: baseline, community-based, and adaptive.
• Decomposed the redistribution problem into sub-components for evolutionary optimization.
• Clustering methods were central to AEA variants, focusing on journey flows and demand adaptation.
• Applied AEA variants to historical data from London's Santander Cycle scheme.

Main Results:

• The AEA variants demonstrated effectiveness in finding solutions for the truck redistribution problem.
• Empirical evaluation using Santander Cycles data validated the AEA's potential.
• The adaptive AEA showed promise in handling dynamic changes in demand.

Conclusions:

• The Artificial Ecosystem Algorithm (AEA) offers a viable heuristic approach for the complex BSS truck redistribution problem.
• AEA's adaptability and decomposition strategy are suitable for distributed computing environments.
• The proposed AEA variants provide a foundation for improving BSS operational efficiency through optimized vehicle balancing.