Article
Computer Science, Artificial Intelligence
Kristijan Jaklinovic, Marko Durasevic, Domagoj Jakobovic
Summary: Scheduling problems are prevalent in various systems, where genetic programming has shown superiority over manual design in generating dispatching rules. Automatically generated dispatching rules perform better than manually adapted rules for constrained problems, showcasing the capability of genetic programming in handling complex real-world constraints.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Automation & Control Systems
Marko Durasevic, Francisco Javier Gil-Gala, Lucija Planinic, Domagoj Jakobovic
Summary: Dynamic scheduling is an important combinatorial optimization problem that frequently occurs in the real world. It is challenging due to its dynamic nature, limiting the effectiveness of improvement-based metaheuristics. This paper proposes a novel collaboration method for ensembles of dispatching rules (DRs) in dynamic scheduling problems, based on a similar method used in static scheduling problems. The results demonstrate that the proposed collaboration method outperforms standard methods, and provide insights for further research in the area of hyper-heuristic ensemble construction.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Marko Durasevic, Domagoj Jakobovic
Summary: Dispatching rules are efficient tools for creating schedules, but manually designing rules for all possible conditions is impractical. Research has focused on automatic design using genetic programming, but evolving rules multiple times may be needed for optimal results. A selection procedure based on problem features has shown to outperform selecting a single rule for all instances.
JOURNAL OF COMPUTATIONAL SCIENCE
(2022)
Article
Computer Science, Information Systems
Lucija Planinic, Hrvoje Backovic, Marko Durasevic, Domagoj Jakobovic
Summary: This paper investigates six different methods for generating dispatching rules and analyzes the results under different scheduling criteria. With the exception of two methods, the performance of the rest is similar, depending on the selected scheduling criterion. Cartesian genetic programming shows the most resistance to bloat issues and evolves dispatching rules with the smallest average size.
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Domagoj Jakobovic
Summary: Scheduling has a significant impact on various areas of human lives. The unrelated parallel machines scheduling problem (UPMSP) is a problem type found in many fields, and there has been an increase in research on this problem in recent years. However, there is currently a lack of systematic overview of the application of heuristic methods for solving the UPMSP. The goal of this study is to provide an extensive literature review on the use of heuristic and metaheuristic methods for solving the UPMSP.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Francisco J. Gil-Gala, Domagoj Jakobovic
Summary: This study investigates the possibility of constructing ensembles of dispatching rules (DRs) to solve multi-objective (MO) scheduling problems. An existing ensemble construction method called SEC is adapted and compared with the NSGA-II and NSGA-III algorithms. The results show that ensembles of DRs achieve better Pareto fronts compared to individual DRs, and SEC achieves equally good or slightly better results than NSGA-II and NSGA-III when constructing ensembles, while being simpler and slightly less computationally expensive.
INTEGRATED COMPUTER-AIDED ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Francisco Javier Gil-Gala, Domagoj Jakobovic, Carlos A. Coello Coello
Summary: Dispatching rules (DRs) are popular methods for solving dynamic scheduling problems, but they perform poorly for multi-objective (MO) problems. Recent research has focused on using genetic programming (GP) to automatically design DRs for MO problems. However, evolving new DRs for each MO problem can be computationally expensive. To address this, we propose a methodology to combine existing DRs for optimizing individual criteria into ensembles suitable for optimizing multiple criteria simultaneously. The method outperforms standard MO algorithms in terms of performance and can be applied to problems with a smaller number of criteria.
SWARM AND EVOLUTIONARY COMPUTATION
(2023)
Article
Management
Cristiane Ferreira, Goncalo Figueira, Pedro Amorim
Summary: The emergence of Industry 4.0 is making production systems more flexible and dynamic, requiring real-time scheduling adaptation. Machine learning methods have been developed to improve scheduling rules, but they often lack interpretability and generalization. This paper proposes a novel approach that combines machine learning with domain problem reasoning to guide the empirical search for effective and interpretable dispatching rules. The experimental results show that the proposed approach outperforms existing literature in various scenarios, indicating its potential as a new paradigm for applying machine learning to dynamic optimization problems.
OMEGA-INTERNATIONAL JOURNAL OF MANAGEMENT SCIENCE
(2022)
Article
Engineering, Chemical
Adilanmu Sitahong, Yiping Yuan, Ming Li, Junyan Ma, Zhiyong Ba, Yongxin Lu
Summary: This study combines genetic programming with feature selection to design effective and interpretable dispatching rules for dynamic job shop scheduling. A new genetic programming method is proposed, which achieves a progressive transition from exploration to exploitation based on population diversity. The experiments show that the proposed approach outperforms other genetic programming-based algorithms in generating more interpretable and effective rules.
Article
Engineering, Industrial
Sungbum Jun, Seokcheon Lee
Summary: This paper proposes a decision-tree-based approach with genetic programming to learn dispatching rules from existing schedules and improve them to minimize delays. Experimental results show that the proposed method outperforms existing rules and provides understandable scheduling insights.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2021)
Article
Engineering, Industrial
Salama Shady, Toshiya Kaihara, Nobutada Fujii, Daisuke Kokuryo
Summary: Thanks to advances in computational power and machine learning algorithms, Genetic Programming (GP) can be used to automatically design scheduling rules for dynamic job shop scheduling problems. However, the computational costs and interpretability of the rules remain significant limitations. In this paper, a new representation of GP rules and an adaptive feature selection mechanism are proposed to improve solution quality by limiting the search space and generating more interpretable rules.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2022)
Article
Mathematics, Applied
Nodari Vakhania
Summary: This paper addresses a basic preemptive scheduling problem where non-simultaneously released jobs are processed by unrelated parallel machines. A fast algorithm is proposed that finds an optimal schedule when the Linear Programming (LP) solution has a small number of non-zero elements. Another linear program is introduced for non-simultaneously released jobs, and an optimal schedule can be constructed from the optimal solution to this linear program. The importance of the procedure lies in the fact that there may exist no optimal schedule that agrees with an optimal LP-solution.
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Domagoj Jakobovic
Summary: Dispatching rules (DRs) are heuristic methods used to solve scheduling problems, typically designed for dynamic conditions, while metaheuristic methods are used for static problems. DRs execute faster and adapt to dynamic changes in the system, making them suitable for situations where quick scheduling or potential system changes are required.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Computer Science, Interdisciplinary Applications
Cristiane Ferreira, Goncalo Figueira, Pedro Amorim, Alexandre Pigatti
Summary: Optimizing operations in bulk cargo ports is crucial due to their importance in international trade. This study focuses on the scheduling problem of unloading wagons in the stockyard, addressing both the deterministic and stochastic versions. Various solution approaches, including Mixed Integer Programming, Constraint Programming, and Genetic Programming, are compared and validated using real data from a leading mining company. The results show that the new heuristic algorithm achieves similar performance to the existing algorithm with significantly reduced computational time. Additionally, the study compares different scheduling strategies and concludes that frequent re-scheduling is the most effective approach in dealing with disruptions while evolved dispatching rules result in fewer deviations from the original schedule.
COMPUTERS & OPERATIONS RESEARCH
(2023)
Article
Engineering, Multidisciplinary
Min Hu, Zhimin Chen, Yuan Xia, Liping Zhang, Qiuhua Tang
Summary: The multi-skill resource-constrained project scheduling problem (MS-RCPSP) is an important management science problem that extends from the resource-constrained project scheduling problem (RCPSP) and is integrated with a real project and production environment. To solve MS-RCPSP, it is efficient to use dispatching rules combined with a parallel scheduling mechanism to generate a scheduling scheme. This paper proposes an improved gene expression programming (IGEP) approach to explore newly dispatching rules that can broadly solve MS-RCPSP. IGEP applies backward traversal decoding mechanism and several neighborhood operators to improve the algorithm's performance. Experiment results show that IGEP discovers ten newly dispatching rules, among which eight outperform other typical dispatching rules.
CMES-COMPUTER MODELING IN ENGINEERING & SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Domagoj Jakobovic
GENETIC PROGRAMMING AND EVOLVABLE MACHINES
(2018)
Article
Computer Science, Artificial Intelligence
Stjepan Picek, Carlos A. Coello Coello, Domagoj Jakobovic, Nele Mentens
JOURNAL OF HEURISTICS
(2018)
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Domagoj Jakobovic
EXPERT SYSTEMS WITH APPLICATIONS
(2018)
Article
Computer Science, Theory & Methods
Mateja Dumic, Dominik Sisejkovic, Rebeka Coric, Domagoj Jakobovic
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2018)
Article
Computer Science, Theory & Methods
Luca Mariot, Stjepan Picek, Alberto Leporati, Domagoj Jakobovic
CRYPTOGRAPHY AND COMMUNICATIONS-DISCRETE-STRUCTURES BOOLEAN FUNCTIONS AND SEQUENCES
(2019)
Article
Computer Science, Artificial Intelligence
Marko Durasevic, Domagoj Jakobovic
JOURNAL OF HEURISTICS
(2019)
Article
Computer Science, Interdisciplinary Applications
Ivan Vlasic, Marko Durasevic, Domagoj Jakobovic
COMPUTERS & INDUSTRIAL ENGINEERING
(2019)
Article
Chemistry, Analytical
Nikolina Frid, Vlado Sruk, Domagoj Jakobovic
Summary: This paper presents five new algorithms for the design space exploration of platforms with sparse connectivity. By leveraging the NSGA-II meta-heuristic and improving the existing mapping algorithm, the chance of finding feasible solutions on such platforms is increased. The authors also propose a synthetic benchmark for further research on these platforms. Experimental results show that the proposed algorithms achieve a high success rate on platforms with dedicated clusters and moderate success rate on tile-like platforms.
Article
Computer Science, Artificial Intelligence
Domagoj Jakobovic, Marko Durasevic, Karla Brkic, Juraj Fosin, Tonci Caric, Davor Davidovic
Summary: Many real-world applications of the vehicle routing problem (VRP) require fast algorithms to generate solutions of acceptable quality for large scale problem instances. The basis for many VRP approaches is a heuristic that builds a candidate solution, which can be improved by a local search procedure. Customised heuristics are needed for specific problem variants in highly dynamic environments, where future information may be uncertain or subject to change.
Proceedings Paper
Computer Science, Artificial Intelligence
Domagoj Jakobovic, Stjepan Picek, Marcella S. R. Martins, Markus Wagner
PROCEEDINGS OF THE 2019 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE (GECCO'19)
(2019)
Proceedings Paper
Mathematics, Interdisciplinary Applications
Stjepan Picek, Domagoj Jakobovic
PROCEEDINGS OF THE 2019 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE COMPANION (GECCCO'19 COMPANION)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Luca Mariot, Stjepan Picek, Domagoj Jakobovic, Alberto Leporati
PARALLEL PROBLEM SOLVING FROM NATURE - PPSN XV, PT I
(2018)
Proceedings Paper
Computer Science, Artificial Intelligence
Stjepan Picek, Karlo Knezevic, Domagoj Jakobovic, Claude Carlet
2018 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC)
(2018)
Proceedings Paper
Computer Science, Artificial Intelligence
Stjepan Picek, Karlo Knezevic, Luca Mariot, Domagoj Jakobovic, Alberto Leporati
2018 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC)
(2018)