Agent-based Scheduling in a Job Shop Using Idle Time and Utilization Information

The internet of things enables communication between the machines located on a shop floor. Machines can become agents which can negotiate with each other and decide in a decentralized fashion which job to process next. This paper studies negotiation rules that can be used in that kind of situation. First, the paper reviews the literature on negotiation rules. Surprisingly, it appears that rules for deciding what to do next do not appear in the literature in a detailed form. Second, by taking the previous gap into account, the paper designs a decentralized scheduling system for job shop systems where jobs have different weights and the total weighted flow time and makespan of the schedule should be minimized. It implements two new negotiation rules that are based on the idle times and remaining processing times of preceding or following machines. An agent-based simulation environment is constructed in order to compare the performances of negotiation rules with a simple priority-weighted shortest processing time rule. The negotiation rules that are studied clearly improve the performance of the system. In practice, the results suggest how the negotiation method that is used should be selected.