A Hybrid Intelligence Approach to Energy Optimization of Waste-Sorting Robots and Its Contribution to Regional Carbon Neutrality

This study proposes a multi-objective optimization framework integrating LSTM-Transformer hybrid models and NSGA-II algorithms to address the energy-efficiency dilemma in waste-sorting robots. The optimized system reduces energy consumption by $15.2 \%$ (average) while maintaining $93 \%$ sorting accuracy, achieving a carbon reduction intensity of 0.48 $\mathrm{kgCO}_{2} / \mathrm{kg}$-waste in pilot cities. By establishing a dynamic carbon accounting model that correlates regional grid emission factors with robotic operational data, we reveal that large-scale deployment in high-emission provinces (e.g., Shandong) could offset $3.2 \%$ of municipal carbon budgets annually. These findings provide actionable insights for intelligent waste management systems under carbon neutrality constraints.