Multithreading MAS Platform for Real-Time Scheduling

Introduction

Scheduling is an optimal resources allocation in order to achieve the set goals. It is an activity, associated with setting the goals and actions in the future.

At the same time, in mathematical terms the scheduling can be considered as a function, where one of the variables is time (Ravindran, 2005). This approach is widely used in modern.

Generally, scheduling implicates the performance of the following stages:

• •

  Statement of problems and goals;

• •

  Compilation of consequence of actions to reach the goals;

• •

  Identification of required resources and sources;

• •

  Identification of the executors and providing them with the schedule;

• •

  Correction of the schedule and execution management (Rzevski, 2014).

As previously mentioned, scheduling is applied in almost every part of the human life, including the production, enterprise, public transport and transportation process organization. Management, traceability of the results become crucial factors both in software and production (Kim, 2015).

Organization of production with hundreds and thousands of people is one of the intractable problems today. The enterprise has to put all of the organization employees’ efforts towards the reaching common goals. Production scheduling in the modern conditions of severe competition in all the economy branches, is not a first step for management to ensure the company efficiency. A properly made production schedule can help to consider the best opportunities of the entire production process.

The main result of scheduling process is a plan, which quality can be measured by the following criteria:

• •

  Schedule flexibility and adaptability;

• •

  Orders delivered on time;

• •

  Resource utilization;

• •

  Response to the unexpected events in real-time;

• •

  Minimum down-time;

• •

  Identification and elimination of production “bottlenecks”.

There are five common methods applied to the production systems scheduling and simulation:

• 1.

  Traditional optimization methods, based on the centralized algorithms such as mathematical methods, used in a compilation of integer and linear programming;

• 2.

  Genetic and neural network scheduling algorithms;

• 3.

  Heuristics algorithms that use business rules;

• 4.

  Multi-agent approaches using algorithms of solving optimization problem in distributed systems with constraints;

• 5.

  Multi-agent approaches using particle swarm optimization (Shi, 2003);

• 6.

  Multi-agent market methods using virtual currencies.

In the next chapters we will provide a brief overview of these methods, focusing on multi-agent approach, which is the basis of the scheduling system that is described in the paper.

Traditional Optimization Methods And Systems

Traditional approach to solving the optimization problems, based on the classical mathematic scheduling methods application, uses invariable and known in advance information about orders, resources and decision-making criteria. Common criteria system does not allow to consider individual preferences, constraints and features of all the resources. Application of this approach does not allow the system to response to the external data changes, which indicates a lack of efficiency of this method for solving complex real-life problems. Moreover, the algorithms of traditional mathematical approaches are rather complex and require large computational resources, which also prevents them from application for quick solution of complex problems under the conditions of daily enterprises operation (Pěchouček, 2008). Companies that apply these methods in their developments, create additional modules for real-life events processing that use hard-coded algorithms. However, these algorithms do not allow to follow the changes and rescheduling after the incoming events.