VR Manufacturing Systems Resource Planning

A Virtual Reality Prototype for Multiagent Based Manufacturing Systems Resource Planning

Research work by Dr. Rashid Jayousi

In today’s ever changing mass customized environment in global market, modern manufacturing systems are required to become more dynamic and adaptive in order to remain competitive[1]. It is found that many industries are changing from mass production to mass customization[5] , therefore manufacturing systems are required to deal with unpredictable demand of different products in small batches. Thus a key factor of success in such environment is the access to a sharable and global view of the enterprise project, product, and services. In recent years researchers have been applying concepts in multi-agent systems to manufacturing systems and have developed new type of manufacturing systems such as Holonic Manufacturing Systems (HMS) [2]. In such manufacturing systems agents cooperate together to manufacture products. We may assume a coordinator agent has a global task (joint task) to create a product with the help of a set of Assembler holons as cohorts, each cohort scheduling its local component of the global task in an environment where some of the scheduling slots (e.g. time-slots) of each cohort will have been already occupied by previously allocated tasks from other coordinators. From the perspective of the coordinators, the ideal situation is when all the required Assemblers are idle, so that the waiting time in between the Assemblers can be reduced to zero, while from the perspective of an Assembler there should not have any idle time at all, so that it does not lose say financially. Thus contentions are inevitable, there is no global optimum, only negotiated compromises. Recognizing this reality, we have developed a model based on user-defined preferences[3], which are expressed on resources used in task scheduling, such as machines, time or labour. It is not possible to meet all the preferences due to contention, and therefore we can define the best solution as a valid solution that meets as many preferences as theoretically possible, a valid solution being a solution that meets all the constraints. Since a this theoretical best is not really practicable, we opt for a good solution that lies within the upper and lower bounds of theoretical predictions. This approach leads to a preference network, where the allocation to a node (i.e. a subtask) affects all other nodes (subtasks) that have preference on that resource. The cascading effect of pre-emption and reallocation leads to branch explosions.

As virtual worlds offer today a new and efficient environment for the multiagent based simulation[4], we propose the use of virtual reality system that can be used to demonstrate our model behavior described. For the simulation study, a manufacturing environment is conjectured, where a set of tasks are resolved into subtasks by coordinator agents, and then these subtasks are allocated to assembler agents through cooperation and negotiation, in which preferred resources are exchanged against payments. In addition this virtual reality simulation can also be used for teaching purposes as it can illustrate the various concepts of multiagent systems. (Ongoing research work)