An agent based and ant colony metaheuristic approach to the last mile logistics problem

Open Access
Kaul, Chaitanya
Graduate Program:
Industrial Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 10, 2018
Committee Members:
  • Daniel Antion Finke, Thesis Advisor
  • Antcolony
  • agentbasedmodelling
  • metaheuristics
  • optimization
  • simulation
  • lastmilelogistics
The last mile logistics refers to the last leg of delivery in a supply-chain transportation network. More transporters and delivery runs are required to fulfil this demand and hence make the last-mile logistic a highly inefficient and time-consuming segment of a supply chain. In this thesis, the formulation of the last mile logistics problem has been presented as an optimization problem and is solved by a modified meta-heuristic algorithm and an agent-based simulation technique. For the last-mile logistics problem, the widely studied vehicle routing problem with time windows transportation model was considered, as it abstracts the salient features of numerous logistics and transportation related real-world problems. The ant colony metaheuristic was then modified to find global minimum in case of the last mile problem. In this work, the last mile logistics problem was then also solved as an agent-based simulation model. Due to its efficient behavioral and communicative patterns, the agent-based systems provide a powerful alternative to traditional optimization techniques. In this work, we present a formulation of the last mile logistics problem as an ant colony metaheuristic and as a multi-agent optimization model. The modified ant colony algorithm was experimented on test problems and data, and successful results were obtained. A detailed experimental assessment of the agent based simulation model and our modified ant colony metaheuristic is presented, including the comparison to the traditional centralized algorithms. A detailed analysis of the solution approach is performed as well, generating future research opportunities. The outcomes of this thesis demonstrate that the methodologies return the best-known solutions accomplished by the state-of-the-art algorithms in most of our experimental cases, representing a significant improvement over the previous comparable agent based or metaheuristic studies. The core commitment of this thesis is the more profound and novel understanding of the implications of embracing agent-based and a heuristic approach to solve the last mile network problems and complex transportation-optimization problems.