The rapid deployment of autonomous haulage systems (AHSs) in open-pit mining has significantly altered haul road geometric design requirements, as autonomous trucks operate under strict kinematic constraints related to turning radius, gradient, and braking performance. Since haulage accounts for 50–60% of total mining costs, optimizing haul road geometry is critical for improving operational efficiency, energy consumption, and safety. This study presents a systematic review of 50 highly relevant studies selected from 81 candidate publications published between 2003 and 2025 through structured database searches and citation chaining. The review synthesizes current developments in haul road layout optimization, turning radius accommodation, gradient design, and safety integration for autonomous mining systems. The findings indicate that GIS-based and integrated optimization approaches consistently improve haulage performance, with reported productivity gains of 5–20%. Turning radius constraints emerged as the primary factor governing kinematic feasibility, while Hybrid A* and its advanced variants represent the dominant path-planning approaches. Recommended gradient limits of 8–12% remain important for balancing efficiency and safety, although emerging AHS-specific models suggest opportunities for controlled relaxation. The review identifies key research gaps in adaptive road design, integrated safety–geometry optimization, and field validation, providing a consolidated foundation for future AHS-compatible haul road design research.

Open-pit mine production scheduling is a complex optimization problem that requires balancing economic performance with operational feasibility and environmental responsibility. While significant advances have been made in mathematical formulations, many existing approaches remain computationally demanding, lack transparency, or are implemented within proprietary systems that limit reproducibility. This study presents a fully reproducible, open-source Python-based optimization framework for open-pit production scheduling that integrates sustainability considerations directly into decision-making. The model employs a mixed-integer “by” formulation with three-dimensional cone-based precedence constraints and incorporates environmental penalties into block valuations to enable ESG-aligned optimization. The framework is designed to be dataset-agnostic and extensible, supporting benchmarking and adaptation to different mining scenarios. The approach is demonstrated using a synthetic block model of 602 blocks over a 12-period planning horizon. The optimization achieved a discounted Net Present Value (NPV) of approximately USD 674.83 million while strictly satisfying precedence, capacity, and operational constraints. Results show that the integration of environmental penalties influences block selection and extraction sequencing, enabling more sustainable scheduling outcomes with only a moderate reduction in economic performance. Sensitivity analyses further confirm the robustness of the framework under varying economic and operational conditions. The study demonstrates that open-source optimization can provide a transparent, adaptable, and effective alternative to proprietary mine planning tools. By combining reproducibility, sustainability integration, and computational efficiency, the proposed framework establishes a practical baseline for future research and large-scale applications in sustainable mine planning.

Open-pit mine production scheduling plays a critical role in maximizing economic returns while meeting environmental and operational constraints. However, the widespread use of proprietary software and rigid deterministic frameworks often limits accessibility, scalability, and adaptability; particularly for small and medium scale operations. This study presents a scalable and sustainable production scheduling model developed entirely in Python, using open-source libraries to bridge this gap. The model employs linear integer programming to optimize the extraction sequence of ore blocks, maximizing Net Present Value (NPV) while respecting precedence relationships, production capacity limits, and environmental penalties embedded in block valuations. Using synthetic block model data, the framework achieved simulated returns exceeding $1.6 billion USD, with integrated visualization tools including 3D diagrams, histograms, and boxplots; enhancing result interpretation. The model's modular and transparent design allows easy adaptation to different mine contexts and planning scenarios. By eliminating reliance on costly platforms and embedding sustainability into the optimization process, this work contributes a practical, technically rigorous, and accessible alternative for modern mine planning. It supports the broader transition toward intelligent, responsible, and inclusive scheduling systems aligned with environmental, social, and governance (ESG) expectations.

Accurate prediction of gold enrichment is critical for mineral exploration and resource evaluation, particularly in data-limited environments where only geochemical information is available. This study evaluates machine learning (ML) models against linear baselines for predicting relative gold enrichment defined as ln(Au/G10), where G10 represents the geometric mean of ten predictor elements (S, Fe, Al, Si, Mn, Sr, Ni, Cu, K, Ti). A total of 53,126 samples from a sulphide-hosted orogenic gold system were subjected to compositional data analysis (CoDA) preprocessing, including multiplicative replacement of below-detection values, closure to a constant sum, centred log-ratio (CLR) transformation of predictor variables, and robust outlier filtering using the Minimum Covariance Determinant (MCD) method. After screening, 41,626 samples were retained for modelling. Comparative modelling included linear baselines—Ordinary Least Squares (OLS), Ridge, Lasso, and Huber—and non-linear ML algorithms: Random Forest (RF), Support Vector Regression (SVR), k-Nearest Neighbours (kNN), and Multi-Layer Perceptron (MLP). Under the non-circular formulation, nonlinear models consistently outperformed linear baselines. Random Forest achieved the strongest validation performance (R2 ≈ 0.51), followed closely by MLP and SVR, while linear models remained substantially weaker (R2 ≈ 0.31). SHapley Additive exPlanations (SHAP) applied to the Random Forest model identified sulphur (S) and iron (Fe) as the most influential predictors, consistent with sulphide-controlled gold mineralization processes in orogenic systems. The model predicts relative gold enrichment based solely on multielement geochemistry, providing a robust and interpretable proxy for mineralization intensity in data-constrained environments. This study demonstrates that machine learning models, when combined with CoDA-correct preprocessing and a non-circular target formulation, can provide geologically meaningful and methodologically robust predictions of gold enrichment. The workflow offers a transparent and reproducible framework for early-stage exploration targeting using multielement geochemistry.

Haul trucks generate more than half of direct greenhouse-gas emissions in typical mining operations, making haulage decarbonization central to the industry’s net zero commitments. Artificial intelligence (AI) is increasingly applied to haulage, but existing reviews examine individual technologies rather than linking them to decarbonization and ESG outcomes. This comprehensive review synthesizes 73 peer-reviewed sources from 2017 to 2026 across route optimization, electric and hybrid trucks, predictive maintenance, fleet monitoring, carbon footprint reduction, and ESG performance, grades reported outcomes by validation basis, and develops a framework linking AI-enabled technology domains, through their operational mechanisms, to ESG outcomes. Reinforcement-learning dispatching reduced haulage emissions by more than 30% relative to fixed scheduling in simulation, trolley assist reduced fuel use by up to 91%, and fleet electrification modeling projected GHG reductions of up to 92.6% in a hypothetical case, contingent on grid carbon intensity. Most reported reductions derive from simulation, and AI’s advantage over conventional control of electrified fleets remains undemonstrated. Gaps include scarce field-verified emission baselines, limited modeling of operator behavior, immature cybersecurity and data governance practice, and weak engagement between ESG research and haulage technology. AI-enabled operational measures offer a promising near-term decarbonization pathway that complements, rather than awaits, fleet electrification.

Mining remains highly energy-intensive. Most sites still rely on diesel because of remote locations and continuous power needs. This dependence is expected to increase as mineral demand rises and ore grades decline. The sector contributes 4–7% of global greenhouse gas emissions. Research on renewable energy integration in mining has expanded but remains fragmented across power systems, mining engineering, life cycle assessment, and social science. This review synthesizes 120 studies spanning solar PV, onshore wind, hybrid microgrids, hydrogen and fuel cells, battery and mechanical storage, and haulage electrification. The review makes three contributions. First, it links operational sustainability outcomes to the upstream social and environmental impacts of mineral supply chains, a connection largely absent from the reviewed literature. Second, it presents a stratified maturity assessment of eleven technology families; solar PV, wind, and electrified haulage prove mature, while hydrogen, redox flow batteries, and pumped hydro remain largely confined to Arctic case studies. Third, it synthesizes cost and emissions outcomes across case studies, identifying a consistent trade-off between low-cost partial diesel offset and high-cost full site decarbonization. The review closes by defining the research frontier: integrated multi-source system design, long-duration storage validation, and unified operational and supply chain sustainability assessment.