Manufacturing activities and production control are constantly growing. Despite this, it is necessary to improve the increasing variety of scheduling and layout adjustments for dynamic and flexible responses in volatile environments with disruptions or failures. Faced with the lack of realistic and practical manufacturing scenarios, this approach allows simulating and solving the problem of job shop scheduling on a production system by taking advantage of genetic algorithm and particle swarm optimization algorithm combined with the flexibility and robustness of a multi-agent system and dynamic rescheduling alternatives. Therefore, this hybrid decision support system intends to obtain optimized solutions and enable humans to interact with the system to properly adjust priorities or refine setups or solutions, in an interactive and user-friendly way. The system allows to evaluate the optimization performance of each one of the algorithms proposed, as well as to obtain decentralization in responsiveness and dynamic decisions for rescheduling due to the occurance of unexpected events.

Production scheduling is an optimizing problem that can contribute strongly to the competitive capacity of companies producing goods and services. A way to promote the survival and the sustainability of the organizations in this upcoming era of Industry 4.0 (I4.0) is the efficient use of the resources. A complete failure to stage tasks properly can easily lead to a waste of time and resources, which could result in a low level of productivity and high monetary losses. In view of the above, it is essential to analyse and continuously develop new models of production scheduling. This paper intends to present an I4.0 oriented decision support tool to the dynamic scheduling. After a fist solution has been generated, the developed prototype has the ability to create new solutions as tasks leave the system and new ones arrive, in order to minimize a certain measure of performance. Using a single machine environment, the proposed prototype was validated in an in-depth computational study through several instances of dynamic problems with stochastic characteristics. Moreover, a more robust analysis was done, which demonstrated that there is statistical evidence that the proposed prototype performance is better than single method of scheduling and proved the effectiveness of the prototype.