Treffer: A machine learning-based genetic programming algorithm for the stochastic resource-constrained project scheduling problem with 3D workspaces and carbon emissions.

Executing activities at different speeds results in varying workspace requirements and carbon emissions. Appropriate workspace allocation is crucial for improving the performance of projects with limited construction sites. Furthermore, uncertainties in activity workloads and weather conditions can significantly impact both project costs and emissions. To address these challenges, we propose a novel stochastic resource-constrained project scheduling problem with 3D workspaces and carbon emissions (SRCPSP-3DWCE), which involves integrated decisions on each activity’s start time, execution speed and workspace location, under spatial constraints and a carbon cap-and-trade mechanism. We formulate an integer programming model for the SRCPSP-3DWCE and develop a machine learning-based genetic programming (MLBGP) algorithm to evolve effective priority rules. The MLBGP integrates two machine learning techniques: a hybrid graph convolutional network – multilayer perceptron (GCN-MLP) model to predict individual fitness and accelerate evolution, and a proximal policy optimisation (PPO) algorithm to adaptively adjust key parameters. The results of numerical experiments and the real-world case study show that MLBGP outperforms 99 combinations of manually designed priority rules, and that the integration of two learning techniques improves both solution quality and computational efficiency. Moreover, considering uncertainties in weather conditions and activity workloads can reduce the misestimation of project makespan and carbon emissions.<bold>Highlights</bold>A new stochastic RCPSP variant is proposed, integrating carbon trading, project scheduling and 3D workspace allocation.Uncertainties in both activity workloads and weather conditions are considered.A machine learning-based genetic programming algorithm is developed to evolve better priority rules.A hybrid graph convolutional network-multilayer perceptron model is designed to predict individual fitness values to reduce computation time.A deep reinforcement learning method is employed to guide the search direction of the algorithm. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of Production Research is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)