Scholarlink

Accurate forecasting of solid waste (SW) generation by type, such as recyclable, contaminated, and general waste, is critical for efficient planning of disposal sites and optimization of collection routes. Existing approaches often fail to capture complex temporal patterns and interdependencies in waste generation, limiting their effectiveness in energy-efficient routing. To address this gap, we develop a data-driven framework that integrates univariate time-series prediction models, including Long Short-Term Memory (LSTM) networks and TimeGPT, a generative pretrained transformer, into a constraint-learning-based optimization model. The framework constructs demand constraints from these forecasts to minimize route energy consumption while also accounting for transportation costs. Applied to the 2020–2021 Austin, Texas dataset, our approach demonstrates improved allocation of waste by type and reduced both cost and energy usage. This study provides a novel methodology for integrating advanced time-series forecasting with route optimization, offering a scalable solution for urban waste management.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Constraint Learning Approach to Route Energy Minimization Using LSTM and TimeGPT-Based Demand Prediction in Waste Management System

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vigneshwar Pesaru

摘要

Accurate forecasting of solid waste (SW) generation by type, such as recyclable, contaminated, and general waste, is critical for efficient planning of disposal sites and optimization of collection routes. Existing approaches often fail to capture complex temporal patterns and interdependencies in waste generation, limiting their effectiveness in energy-efficient routing. To address this gap, we develop a data-driven framework that integrates univariate time-series prediction models, including Long Short-Term Memory (LSTM) networks and TimeGPT, a generative pretrained transformer, into a constraint-learning-based optimization model. The framework constructs demand constraints from these forecasts to minimize route energy consumption while also accounting for transportation costs. Applied to the 2020–2021 Austin, Texas dataset, our approach demonstrates improved allocation of waste by type and reduced both cost and energy usage. This study provides a novel methodology for integrating advanced time-series forecasting with route optimization, offering a scalable solution for urban waste management.