Dynamic Multi-agent Real Time Scheduling Framework for Production Management

Production management is a complex problem. The problem becomes more complex in open, uncertain, dynamic environments and distributed productions. The existing static scheduling methods are not efficient to deal with this problem efficiently. This paper proposed multi-agent approach for real time scheduling in production management to efficiently manage production. The proposed scheduling optimizes the performance of production management in dynamic uncertain environment. Dynamic multi-agent approach in real time scheduling consist of the agent collect and analysis the production data in real time and ensure real time response to emergency events. Real time scheduling agents increase the resource utilization and success rate in production management by using real time feedback.


Introduction
In production management, different manufacturing resources are scheduled to manufacturing processes with time slots in plan. The production environment is open, uncertain and dynamic in nature. To cope with such an uncertain production environment dynamic scheduling of manufacturing resources is required. The production scheduling is challenging problem particularly in open, dynamic and uncertain environment where the resources may change in time, some resources can be added into the manufacturing system run time or removed from the system. The tasks also change its parameter run time. The existing centralized production management is mainly based on material requirement planning and enterprise resource planning. Theses existing production management scheduling techniques having many limitations as follows. 1. The current approach not efficient to handle uncertain events 2. Not efficiently deals with production disturbances such as failure of machine, late supply of materials and unavailability of employee 3. Existing schedule not efficient to handle run time changes in the manufacturing processes 4. Optimal utilization of resources not possible in the conventional methods. 5. Current production management methods are incorrect, inconsistent, and incomplete and produce time delay. 6. In production management methods not based on dynamic real time response and adaptive control. To address theses problem this paper proposed multi-agent approach in real time scheduling for efficient production management. Multi-agent approach in real time scheduling increase the resource utilization and the reliability of resources. The real time scheduling monitors the performance of shop floor and react with the changes in real time.

Background Work
Multi agent technology recently used in production management for optimization. Different multi agent based production management techniques are described in the past [1]. Dynamic scheduling approach is used to manage the production demands in real time [2]. The accuracy of the thesis system is based not only on correct output but also the time when the output produced [3]. The dynamic scheduling problem is of two type one is dynamic planning and dynamic best effort approach [4]. In a dynamic planning approach the job is scheduled only when the deadline can be achieved and in dynamic best effort scheduling job can schedule without checking for deadline guarantee [5]. The dynamic scheduling also categories into two types based on hardness of deadline [6]. In hard dynamic scheduling the job deadline is crucial and hazardous if the deadline of the job is not achieved. Soft dynamic scheduling systems and deadlines are not crucial. The production management is dynamic hard scheduling with dynamic planning approach. Recently multi agent based technology used intelligent dynamic scheduling in production management in factory automation. Multi agent scheduling with contract negotiation protocol which is used to solve many scheduling problems with failure detection. Decentralized manufacturing using multi agent is used to address dynamic resource sharing. Real time scheduling using a multi agent system is presented for transportation problems. Distributed agent based scheduling based approach is presented for a flexible manufacturing system [7]. Radio frequency based identification system used with multi agent system for tracking and tracing manufacturing problems. Some authors also presented a simulation based approach to address internal disturbances in the manufacturing process [8] [9]. Memory aware resource scheduling algorithm is proposed for energy efficient cloud data centers [10].

Dynamic Real Time Multi-agent Scheduling Framework
In a dynamic multi-agent approach in real time scheduling framework, the agents perceive the environment and the status of the production system [11]. Environment in production management consists of job arrival, machine computing capacity, resource allocation etc. The physical system in production management is represented as machine and system state. The system state represents the time required to process the unit processing, batches to be processed, length of queue and the critical information.
The agent action is based on events and falls into operation such as preventive, active, protective and informative as it senses the system as well as machine state and takes dynamic real time decision and adaptive control. The real time multi-agent scheduling framework for production management is as shown in Figure 1. The proposed real time multi-agent scheduling framework consists of four agents such as Machine State Agents, Machine Status Evaluation Agent, Production Resource Monitoring Agent and Dynamic Real Time Scheduling Agent.

Machine Status Agent
This agent is responsible for collection of real time status information of machine different health parameters. The machine status agents collect the different parameters of machines such as machine age, machine working parameter using real time sensing mechanism. The machine status agent also processes raw data of machine into meaningful information which is used for real time application services. The machine status agent consists of two main modules Data Sensing Module: It senses the real time status information from machines through different sensors. The real time operating system enabled sensing mechanism is used to sense the real time status information of the machine. The different sensors used to get the status of a machine are pressure sensor, vibration sensor, RFID, motion sensors etc. Real Time Analytic Module: This module responsible for analysis of real time information captured by a data sensing module.

Machine Status Evaluation Agent (MSEA)
It is responsible for evaluation of machine status information given of machine status agents. This agent evaluates the status information and assigns tasks to the machine for optimal scheduling. It assigns tasks of production management processes to machines as shown in Figure 3. The set of tasks of production processes assigned to the machines as per the machine capacity management unit information and select the optimal machine to which the tasks are assigned. The machine capacity management unit is used to balance load of machines. Machine capacity management unit calculate the machine load by interacting with machine status agents to get the information related current load of the machines and process assigned to machine having less load.

Production Resource Monitoring Agent
Resource management and monitoring is done by production resource management agents. It also tracked production changes, disturbances in manufacturing processes. Production resource monitoring agents act as an interface between the machine status agent and dynamic real time scheduling agent. It collects the real time information from the machine status agent and sends this to dynamic real time scheduling agents. The production resource monitoring agent gets the details about the production order, its schedule etc. The collection of all the machine status agents is collected and sent to dynamic real time scheduling agents.

Figure 4. Production Resource Monitoring Agent
Real time information tracking and integration unit of production resource monitoring agent collect real time information from the machine status tracking agents and integrate it. Date source unit collect production order information and send the information to dynamic real time scheduling agents for taking the optimal scheduling information

Dynamic Real Time Scheduling Agent
It runtime decides the order of execution of tasks of processes in production management. The dynamic real time scheduling agent takes input from machine status evaluation agents and processes monitoring information to schedule to the given tasks. The output of the dynamic real time scheduling agent is that the given process of tasks assigned to the machine within the deadline. Fig.5 shows the dynamic real time scheduling agent which consist of three module problem analysis module, problem solving module and rescheduling module.

Results and Discussion
System model of dynamic real time scheduling is presented in Table 1. The different production management consists of a set of processes of tasks and machines. The deadline of each task is used to ensure that all the tasks must finish its execution before its deadline for real time operation. The objective of dynamic real time scheduling is to complete execution of process of tasks within the  The proposed algorithm calculated the machine load dynamically for each individual process of tasks and assigned process to machines with minimum load so that the execution of process must get finished its operation. The proposed dynamic multi agent real time scheduling framework is used to improve resource utilization as compare to existing methodology. The proposed framework simulation is applied to set of production tasks T={500, 1000,1500,2000,2500,3000,3500,4000,4500,5000}. The resource utilization is measured with the proposed framework and the proposed scheduler increases the resource utilization and success rate.

Conclusions
Production management is challenging problem in uncertain, dynamic, open system and distributed environment. The existing static scheduling approach in production management is not efficient in terms of success rate and resource utilization, The proposed dynamic multi agent real time scheduling framework increase resource utilization and success rate where the overall production management operations is manage and coordinated by different agents such as machine status agent gives status information related with health of machine which is further evaluated by machine status evaluation agent and the information used by production and resource monitoring agent for efficient scheduling of production tasks by dynamic real time scheduling agents.