A new Multi-Agent Architecture for Job Shop Production Scheduling and Control

During the last years the growing requests for ampler ranges of products, greater level of personalization, higher qualitative standards and shortly conditions of delivery, have imposed the necessity to promote Flexible Manufactoring Systems (FMS). An efficient method to favour the management and the control of FMS, is constituted by the Autonomous Agents Systems with which the whole productive organization in considered as composed by autonomous work drives (the agents) with a mechanism of cooperation founded on the logic of the contract-net. Starting by a detailed analysis of the available technical literature we propose a scheme for negotiation among the agents as half view between the part-initiated scheme and the resource-initiated scheme. Such scheme is based on the individualisation of five decisional factors (Stock Agent; Resource Agent; Planner Agent; Buffer Agent; Manager of transport system) referred to a productive plant with job shop layout (five departments, with two engines for department). In this context, we have established, first the principal functions of the agents (availability of engines to operation; sequences of the jobs; planning and management of movements; management of engines failure), and in a second step the targets increasing the efficiency of the productive cycle based on dispatching rules (Shortest Process Time; Earliest Due Date; Minimum Slack Time; Critical ratio), so, finally, we have developed a scheduling algorithm developed by a software (Arena 3.0) able in manufacturing problem solving. The results of the experimentation are shown by graphics (histograms) highlighting how the Autonomous Agents System are able to enhance, in the different created productive situations, the analysed targets of performance. In the end, we noticed that our approach confirm the validity of the architectures multi-agent in the scheduling and control of manufacturing productive system oriented to flexibility.