A heuristic-assisted deep reinforcement learning algorithm for flexible job shop scheduling with transport constraints

This study proposes an approach to minimize the maximum makespan of the integrated scheduling problem in flexible job-shop environments, taking into account conflict-free routing problems. A hybrid genetic algorithm is developed for production scheduling, and the optimal ranges of crossover and mutation probabilities are also discussed. The study applies the proposed algorithm to 82 test problems and demonstrates its superior performance over the Sliding Time Window (STW) heuristic proposed by Bilge and the Genetic Algorithm proposed by Ulusoy (UGA). For conflict-free routing problems of Automated Guided Vehicles (AGVs), the genetic algorithm based on AGV coding is used to study the AGV scheduling problem, and specific solutions are proposed to solve different conflicts. In addition, sensors on the AGVs provide real-time data to ensure that the AGVs can navigate through the environment safely and efficiently without causing any conflicts or collisions with other AGVs or objects in the environment. The Dijkstra algorithm based on a time window is used to calculate the shortest paths for all AGVs. Empirical evidence on the feasibility of the proposed approach is presented in a study of a real flexible job-shop. This approach can provide a highly efficient and accurate scheduling method for manufacturing enterprises.

Motivated by high ecological and economical potentials, remanufacturing is receiving increasing attention as a process that puts used products into an “as-good-as-new-or-better” condition. Here, many challenges occur, which are unseen in manufacturing, resulting from unknown conditions of used products. One example is the stochastic routing of the products through the remanufacturing system (RS), which makes a flexible material transport by e.g. automated guided vehicles (AGVs) necessary. This places special demands on the control of the RS. To handle these uncertainties a decentralized-hierarchical control architecture is presented. The scheduling of machines and AGVs is optimized simultaneously using Constraint Programming (CP). All the participants within the RS will be networked as a cyber-physical system and controlled by decentralized components. The approach is validated through simulation and the implementation in a model factory. Results show a 35.6% makespan reduction, short computing times and the ability to react adequately to unforeseen events.

Multi-objective green scheduling of integrated flexible job shop and automated guided vehicles

This paper addresses machines, automated guided vehicles (AGVs), tool transporter (TT) and tools simultaneous scheduling in a multi-machine flexible manufacturing system (FMS) to minimize makespan (MSN). The transfer times of tools and jobs are considered. Only one copy of each type of tool is made available due to economic restrictions. The tools are stored in a central tool magazine (CTM) that shares with and serves for several machines. Jobs and tools between machines are transferred by AGVs and TT, respectively. This simultaneous scheduling problem is highly complex in nature as it involves job-operations sequencing on machines, assigning AGVs and tools to job-operations and associated trip operations including the times of empty trip and loaded trip of both AGVs and TT. This paper presents nonlinear-mixed integer programming (MIP) formulation to model and symbiotic organisms search (SOS) algorithm based on the symbiotic interaction strategies that organisms use to survive in the ecosystem to solve this simultaneous scheduling problem. The results have been tabulated and analyzed. An industrial problem in a manufacturing company is used for verification and the result obtained by SOS algorithm is compared with other algorithm.

Existing research on flexible job shop scheduling (FJSP) often overlooks critical logistics constraints, particularly the limited availability and multi-load capacity of Automated Guided Vehicles (AGVs), resulting in a significant disconnect between theoretical models and the operational realities of service-oriented smart manufacturing systems. To bridge this gap, this study proposes a Non-dominated Sorting Genetic Algorithm II with Local Search Algorithm (NSGA-II_LS) for green flexible job shops with multi-load AGV coordination (GFJSP-MA), which integrates autonomous optimisation of makespan, machine utilisation, and carbon emissions through intelligent knowledge integration and dynamic operational adaptation. The three core innovations aligned with autonomous manufacturing system design are as follows: (1) Mining the domain knowledge of GFJSP-MA and designing a novel encoding and decoding method; (2) adaptively improving the genetic operators to enhance the global exploration capability; (3) developing eight local search operators based on the problem characteristics to enhance the convergence of the algorithm. This study was tested and validated on a generalised dataset and compared with three other multi-objective algorithms, and the experimental results verified the superiority of NSGA-II_LS for solving GFJSP-MA.

This paper investigates the Dynamic Flexible Job Shop Scheduling Problem (DFJSP), which is based on new job insertion, machine breakdowns, changes in processing time, and considering the state of Automated Guided Vehicles (AGVs). The objective is to minimize the maximum completion time and improve on-time completion rates. To address the continuous production status and learn the most suitable actions (scheduling rules) at each rescheduling point, a Dueling Double Deep Q Network (D3QN) is developed to solve this problem. To improve the quality of the model solutions, a MachineRank algorithm (MR) is proposed, and based on the MR algorithm, seven composite scheduling rules are introduced. These rules aim to select and execute the optimal operation each time an operation is completed or a new disturbance occurs. Additionally, eight general state features are proposed to represent the scheduling status at the rescheduling point. By using continuous state features as the input to the D3QN, state-action values (Q-values) for each scheduling rule can be obtained. Numerical experiments were conducted on a large number of instances with different production configurations, and the results demonstrated the superiority and generality of the D3QN compared to various composite rules, other advanced scheduling rules, and standard Q-learning agents. The effectiveness and rationality of the dynamic scheduling trigger rules were also validated.

The increasing uncertainty in modern manufacturing systems presents challenges in managing disruptions. Unlike traditional linear production systems, matrix production enables modular flexibility, allowing for real-time machine rescheduling and dynamic material transport via Automated Guided Vehicles (AGVs). However, this added flexibility also introduces substantial complexity in decision-making, particularly under disruption scenarios, as it necessitates the simultaneous adjustment of both AGV dispatching and machine scheduling. Traditional optimization- and rule-based approaches to disruption management often suffer from extended decision-making times, suboptimal performance, and, critically, the limited integration of human expertise. To address these limitations, we propose a Generative AI (GenAI)-driven decision support framework that integrates Large Language Models (LLMs) within matrix production systems to enhance manufacturing resilience. Our framework leverages a fine-tuned GenAI model trained on structured knowledge bases derived from digital twin-enabled what-if disruption scenarios and expert knowledge. By integrating human-guided decision refinement, the framework provides real-time, human-in-the-loop, and explainable decision support aligned with operational goals and human preferences. To validate the approach, we implemented the framework in a semiconductor production use case and assessed the system resilience via key metrics such as utilization, downtime, and throughput. Compared to the best baseline static rule, our GenAI-based framework improved throughput by 11.81% and machine utilization by 11.49% in disruption scenarios. The results show that task-specific recommendations generated by the GenAI model effectively assist workers in selecting optimal response decisions during real-time disruptions.

Multi-objective optimization for energy-efficient flow shop scheduling problem with blocking and collision-free transportation constraints

In the context of Industry 5.0, worker well-being is a central tenet within the production process. This research work aims to contribute to this goal by integrating ergonomic factors into task allocation and scheduling while maintaining high productivity. An ergonomics-aware operator-task allocation and scheduling algorithmic framework is developed for a flexible job shop production system to minimise the makespan while ensuring that the maximum physical burden on each operator is not exceeded. The proposed algorithmic framework can integrate any method that assesses the composite ergonomic load on an operator who has been assigned multiple tasks. It can handle several ergonomic measurements of different risk factors with their predefined threshold to ensure safety and efficiency. This flexibility enables decision-makers to select the best measures for their needs. This framework is tested on small to moderate-sized instances, and the numerical experiments show that the proposed algorithmic framework converges to a solution satisfying the ergonomic risk requirement, at the expense of only a slight increase in the makespan. The expected reduction in ergonomics-related injuries in companies adopting this framework would result in an actual productivity gain, which may turn out to be significantly higher than the minor loss due to the increased makespan.

Enhanced NSGA-II for multi-objective energy-saving flexible job shop scheduling

We propose an approach for finding an optimal path in a flexible jobshop manufacturing system considering two criteria of time and cost. With rise in demands, advancement in technology and increase in production capacity, the need for more shops persists. Therefore, a flexible jobshop system has more than one shop with the same duty. The difference among shops with the same duty is in their machines with various specifications. A network is configured in which the nodes are considered to be the shops with arcs representing the paths among the shops. An automated guided vehicle functions as a material handling device through the manufacturing network. To account for uncertainty, we consider time to be a triangular fuzzy number and apply an expert system to infer cost. The objective is to find a path minimising both the time and cost criteria, aggregately. Since time and cost have different scales, a normalisation procedure is proposed to remove the scales. The model being bi-objective, the analytical hierarchy process weighing method is applied to construct a single objective. Finally, a dynamic programming approach is presented for computing a shortest path in the network. The efficiency of the proposed approach is illustrated by a numerical example.

To improve the scheduling efficiency of flexible job shops, this paper proposes a multi-objective collaborative intelligent agent reinforcement learning algorithm based on weight distribution. First, a mathematical model for flexible job shop scheduling optimization is established, with the makespan and total energy consumption of the shop as optimization objectives, and a disjunctive-graph is introduced to represent state features. Second, two intelligent agents are designed to address the simultaneous decision making problems of jobs and machines. Encoder-decoder components are implemented to facilitate state recognition and action output by the agents. Reward parameters are computed based on temporal differences at various moments, constructing a multi-objective Markov decision-process training model. Using hypervolume, set coverage and inverted generational distance as evaluation metrics, the algorithm is compared with those proposed in other studies on standard instances. The results demonstrate that the proposed method significantly outperforms other algorithms in solving the flexible job shop scheduling problem. Finally, a real-world case study further validates the effectiveness and practicality of the proposed algorithm.

To study the flexible job-shop problem(FJSP) with considering the transporting time and deteriorating effect, a novel model with the objective of minimizing the makespan is developed by disjunction graph and a linear deteriorating function whereas a hybrid algorithm (IGAVNS) is proposed combining immune genetic algorithm (IGA) with variable neighborhood search strategy (VNS). First, the operation-based encoding method is used to encode chromosome individual, solving the operation sequencing and machine selection. Then, elite individual is searched by the proposed algorithm with five neighborhood structures. Simulations were carried out by different algorithms on the cases of different sizes. The results were compared and analyzed, verifying the effectiveness of the proposed algorithm.

First, a multi-objective immune genetic algorithm integrating immune algorithm and genetic algorithm for flexible job shop scheduling is designed. Second, Markov chain is used to analyze quantitatively its convergence. Third, a simulation experiment of the flexible job shop scheduling is carried out. Running results show that the proposed algorithm can converge to the Pareto frontier quickly and distribute evenly along the Pareto frontier.

PurposeThe purpose of this paper is to focus on the identification of barriers to the implementation of benchmarking initiatives. Managers have little guidance on strategies for successful implementation and face difficulties in anticipating problems. The main objective of this paper is to provide a better understanding of typical barriers that occur in a specific type of benchmarking: internal benchmarking.Design/methodology/approachThis paper systematically develops a categorized list of barriers to internal benchmarking based on the literature; and validates and enriches them with an in‐depth case study of an internal benchmarking initiative.FindingsBarriers to internal benchmarking initiatives of different types are found: organizational barriers (people, culture, and context), benchmarking project management barriers (planning and implementation, leadership, and business pressures) and benchmarking data barriers (difficulty to access/compare data). Future studies might take these categories as a solid starting point in furthering the understanding of barriers to benchmarking.Research limitations/implicationsThe exploratory nature of the paper limits its generalizability.Practical implicationsThe paper paves the way for the process of devising strategies to overcome the identified barriers. This is especially relevant because few studies to date have focused on these issues and benchmarking practitioners have insufficient knowledge to anticipate and overcome benchmarking‐related problems.Originality/valueThis paper contributes to theory building on the implementation of benchmarking in general; and on barriers to internal benchmarking initiatives in particular.