and introducing constraints that prevent underutilization of purchased trucks. The resulting extraction schedule varies significantly from the base case, leading to the pit evolution along the full length of the deposit, progressively deepening along the dip. This, compared to a more segmented base schedule, results in shorter trucking times and less inclined haul roads. Additionally, a decreased amount of waste mined leads to a reduction in fleet requirements. Over the 10-year planning period, the number of necessary hauling trucks is reduced from 35 to 20, resulting in over 20% reduction in capital and operational costs. Although emissions associated with the production of the vehicles are not explicitly mentioned, a 40% reduction in fleet size noticeably decreases the Scope 3 emissions of the project. A more direct approach is presented in De Carvalho and Dimitrakopoulos (2024). The researchers developed a stochastic shovel allocation solution, which updates the daily shovel movement decisions using a reinforcement learning actor-critic algorithm. The short-term schedule model is continuously updated with blasthole data, allowing for accurate grade control and reducing geological uncertainty. While geological uncertainty is modelled based on the initial drilling campaign and updated on a regular basis, the behavior of shovels and comminution equipment is simulated using historical data from the mining operation. The solution was tested in a copper mining complex, comprised of two open pits with a total of 18 shovels. Within the studied 30 day period, the reinforced learning approach outperformed the base schedule by producing substantially more copper, decreasing the quantity of waste sent to the dump, and generating an additional 27% of cash flow. In contrast to the two previously mentioned studies, Kazemi Ashtiani et al. (2025) aims to optimize the fuel consumption of the haul and shovel fleet directly, by minimizing idle and waiting time as well as fuel consumption of the active truck. The uncertainty associated with vehicle performance is modelled by fitting best suited distribution functions for each of the variables based on historical data, and the objective function is solved using CPLEX. The solution is tested by developing a case study set in an open-pit iron mine in Iran, and the results demonstrate a 6% reduction in fuel consumption per ton of material transported over the evaluated 10-day period while maintaining the scheduled production rate. Table 3 – Summary of transport optimization methodologies Author Year Objective function Planning scope Model type Solution methodology/Solver Spleit and Dimitrakopoulos 2017 Maximize NPV, minimize trucking cost, penalize deviations from production targets Longterm SMIP CPLEX, sequential optimization de Carvalho and Dimitrakopoulos 2024 Minimum loss function with processing costs Shortterm Metaheuristics with ML (reinforcement learning) Kazemi Ashtiani et al. 2025 Minimize idle and waiting time, minimize fuel Shortterm SMILP CPLEX
RkJQdWJsaXNoZXIy MTM0Mzk2