Integrated scheduling of jobs, tools, and AGVs in FMS with non-identical machines using a recurrent neural network

In a flexible manufacturing system (FMS), scheduling jobs and tools across non-identical machines, integrating automated guided vehicles (AGVs), and considering multi-objective functions, constitutes a significant obstacle for typical mathematical optimization techniques. Herein, we consider scheduling jobs, tools, and AGVs in an FMS that consists of three non-identical machines. The multi-objective functions targeted are tooling cost minimization and makespan reduction. The non-identical machines' processing rates are specified in the ratio of 1:1.2:1.4. Each of the tools (T1, T2, and T3) is available in a single mode, with T1 being more expensive than T2, which is more expensive than T3. To address such a complex optimization problem, we use a Recurrent Neural Network (RNN) and an Improved version to obtain near-optimum solutions and evaluate such algorithms' comparative performance. The average computation time to determine the optimal sequence was reduced from 10.33 minutes to 6.24 minutes (for a 4-job problem) as we employed the Improved RNN algorithm instead of the RNN algorithm.