The paper discusses the use of data-driven models to predict the outcome of autonomous wheel loading actions. These models utilize deep neural networks trained on over 10,000 simulations to predict the loaded mass, time, work, and resulting pile state of a loading action based on the initial pile state. The aim is to support optimal planning and increase performance and robustness in construction and mining equipment operations. The authors suggest that this approach can be an effective solution to the increasing labor shortage in these industries.
Publication date: 22 Sep 2023
Project Page: https://arxiv.org/abs/2309.12016v1
Paper: https://arxiv.org/pdf/2309.12016
