Idioma:

Optimizing Coordination on Road Construction Sites with a Reinforcement Learning Framework

Bruce, Øystein Høistad

Texto completo disponível

Citações Citado por

Título:
Optimizing Coordination on Road Construction Sites with a Reinforcement Learning Framework
Autor: Bruce, Øystein Høistad
Assuntos: capacitated vehicle routing problem ; CO2 reduction ; fuel consumption model ; graphs ; machine learning ; multi-agent reinforcement learning ; neural networks ; reinforcement learning ; road construction sites
Descrição: Road construction sites are often inefficient, with construction machines frequently idling for extended periods, wasting fuel and time. One way to increase efficiency is to optimize the scheduling of the dumpers transporting materials across the site. This thesis proposes and investigates a multi-agent reinforcement learning framework designed to coordinate dumpers and excavators on construction sites. The framework generates time schedules for all vehicles, considering multiple criteria such as fuel consumption, completion time, and cost. Users can choose a plan that best aligns with their preferences, ensuring maximum efficiency. The framework has a negligible training time and generally outperforms a baseline constructed from human behavior. In addition, we developed a predictive fuel consumption model for a dumper using high-resolution data logged over 12 working days. By associating each dumper with such a model, we can more accurately predict their fuel consumption while driving, further improving the planning.
Data de criação/publicação: 2023
Idioma: Inglês

Links

Voltar para lista de resultados

Anterior Resultado 2 Avançar Ir para próxima página