A multi-agent system for a bus crew rescheduling system

Abstract

Unpredictable events (UE) are major factors that cause disruption to everyday bus operation. In the occurrence of UE, the main resources - crews and vehicles - are affected, and this leads to crew schedule disruption. One way to deal with the problem is crew rescheduling. Most of the current approaches are based on static schedules do not support rescheduling in a real-time scenario. They have the ability to reschedule but a new complete schedule is produced without concerning the real time situation. The mathematical approaches which are used by most scheduling packages have the ability to search for optimum or near optimum schedules but they are usually slow to produce results in real-time because they are computationally intensive when faced with complex situations. In practice, crew or bus rescheduling is managed manually, based on the supervisor's capabilities and experience in managing UE. However, manual rescheduling is complex, prone to error and not optimum, especially when dealing with many UE at the same time. This research proposes the CRSMAS (Crew Rescheduling System with Multi Agent System) approach as an alternative that may help supervisors to make quick rescheduling decisions by automating the crew rescheduling process. A Multi Agent System (MAS) is considered suitable to support this rescheduling because agents can dynamically adapt their behaviour to changing environments and they can find solutions quickly via negotiations and cooperation between them. To evaluate the CRSMAS, two types of experiment are carried out: Single Event and Multiple Events. The Single Event experiment is used to find characteristics of crew schedules that influence the crew rescheduling process while the Multiple Events experiment is used to test the capability of CRSMAS in dealing with numerous events that occur randomly. A wide range of simulation results, based on real-world data, are reported and analysed. Based on the experiment it is concluded that CRSMAS is suitable for automating the crew rescheduling process and capable of quick rescheduling whether facing single events or multiple events at the same time, the success of rescheduling is not only dependant on the tool but also to other factors such as the characteristics of crew schedules and the period of the UE, and one limitation of CRSMAS that was discovered is it cannot simulate different type of events at the same time. This limitation is because in different events there are different rules but, in Virtual World, agents can only negotiate with one set of rules at a time.