Designing a multi-agent system for monitoring and operations recovery for an airline operations control centre

• Autor(a) principal: António Jesus Monteiro de Castro
• Data de Publicação: 2007
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://repositorio-aberto.up.pt/handle/10216/342
• topic: Engenharia de computadores; Computer engineering

An airline schedule seldom operates as planned. Problems related with aircrafts, crew members and passengers are common and the actions towards the solution of these problems are usually known as operations recovery or disruption management. The Airline Operations Control Center (AOCC) tries to solve these problems with the minimum impact in the airline schedule, with the minimum cost and, at the same time, satisfying all the required safety rules. Usually, each problem is treated separately and some tools have been proposed to help in the decision making process of the airline coordinators. We have observed the AOCC of TAP Portugal, the major Portuguese airline, and, from those observations, several hypotheses have been identified and some of them experimented. We believe, and that is one of our main hypothesis, that the Multi-Agent System (MAS) paradigm is more adequate to represent the multi-level hierarchy organization and the several roles that are played in an AOCC. In this thesis we propose the design and partial implementation of a Distributed MAS representing the existing roles in an AOCC. We hypothesize that if we take advantage of the fact that each operational base has specific resources (both crew and aircrafts) and that if we include information regarding costs involved (for example, crew payroll information and hotels costs, among others), the solutions to the detected problems will be faster to find and less expensive. We also hypothesize that if we use specialized software agents that implement different solutions (heuristic and other solutions based in operations research models and artificial intelligence algorithms), to the same problem, the robustness of the system will increase. Finally, we believe that the inclusion of some kind of learning mechanism that learns from previous utilization of crew members will improve the solutions quality. Extending that learning mechanism to learn each crew member profile, and applying that knowledge for generating future schedules, the management of that expensive resource will be much more efficient and the level of satisfaction of each crew member will increase. We also present a real case study taken from TAP Portugal AOCC, where a crew recovery problem is solved using the MAS. Computational results using a real airline schedule are presented, including a comparison with a solution for the same problem found by the human operators in the Airline Operations Control Center. We show that, even for simple problems, and when comparing with solutions found by human operators in the case of this airline company, it is possible to find valid solutions, in less time and with a smaller cost. In this thesis we also show how we complement the GAIA methodology in order to better analyze and design the proposed MAS for the AOCC. Besides showing the rationale behind the analysis, design and implementation of our system, we also present how we mapped the abstractions used in agent-oriented design to specific constructs in JADE. The advantages of using a goal-oriented early requirements analysis and its influence on subsequent phases of analysis and design are also presented. Finally, we also propose UML 2.0 diagrams at several different levels for representation of GAIA deliverables, like organizational structure, role and interaction model, agent and service model.