Multi-Hop Ridematching optimization problem: Intelligent chromosome agent-driven approach

Abstract

In order to solve the now ubiquitous transport problems we face, may they be financial or environmental, we are primarily interested in the development of a dynamic optimized ridesharing service. Shared cars was developed in order to meet transport needs (spatio–temporal flexibility) and to promote the co-modal practice. The focus is thus for different modes of transport, whether public or private, to complement each other by integrating the car pooling system as an efficient alternative. Keeping this in mind, we concentrate on setting up an automatic and optimal ridesharing system. Said system is part of an intelligent co-modal platform which provides the required efficiency in such a context. In order to solve the problem, we must create a Ridematching solution with an arbitrary number of transfers that respects the personal preferences of the users as well as their time constraints. However, as considered to be NP-Complete, a more efficient metaheuristics is required in the application in order to solve the dynamic Multi-Hop Ridematching problem (MHRP). Evolutionary Algorithms (EAs) are known as a powerful and robust optimization technique. Nevertheless, EAs are not only expensive but also difficult to configure. This study puts forward an original Evolutionary Approach in which the chromosomes are defined as Autonomous and Intelligent Agents (E2AIA). Through an accurate protocol negotiation, the chromosomes agents can control the genetic operators and provide guidance when searching for optimal solutions within a reasonable time. This is crucial in real-world systems, where time for deliberation is a very important factor. Test results indicate that the classical evolutionary algorithm, developed to solve the dynamic MHRP, results in poor performance when compared with our E2AIA in terms of optimality and computational time. It is worth noting that the proposed agent-driven method is general and could be adapted to expert design and intelligent systems for other complex search issues.