A Self Adaptive Neural Agent Based Decision Support System for Solving Dynamic Real Time Scheduling Problems

Abstract

Manufacturing production systems are facing growing challenges to ensure competitive advantages and survive world-class under a growing competition and increased customers' requirements. Production scheduling lies at the heart of the manufacturing systems performance, therefore, a better resolution of scheduling problems will ensure the above objectives. Agent based systems were widely adopted in the literature as dynamic and adaptive approaches capable of enhancing scheduling decisions, relying on their distributed nature as well as the adaptive and intelligent behavior of agents. In this study we present a self-adaptive neural-agent-based decision support system where we integrate a set of neural network (NN) candidates at each control system agent. This integration gave rise to simultaneously exploited agents' communication and NNs' learning capabilities. It builds an innovative agent-based architecture able to make suitable real time scheduling decisions, and to improve the production efficiency based on embedded NNs assistance. Thus, according to our experimental results, the self-adaptive neural-agent-based decision support system was able to effectively and efficiently yield better results for production schedule, especially with respect of the mean tardiness of jobs that was significantly reduced when compared to the outcomes of the other applied methods.