A Double-Layer Vehicle Speed Prediction Based on BPNN-LSTM for Off-Road Vehicles

Area of Science:

• Engineering
• Computer Science

Background:

• Accurate vehicle speed prediction (VSP) is vital for energy management, particularly for off-road vehicles where existing methods are limited.
• Off-road vehicle dynamics present unique challenges for VSP, necessitating specialized approaches.

Purpose of the Study:

• To develop and evaluate a novel double-layer VSP method tailored for off-road vehicles.
• To enhance the accuracy and real-time performance of VSP for heavy machinery like mining trucks and loaders.

Main Methods:

• A double-layer framework integrating a long short-term memory (LSTM) network for speed prediction and a backpropagation neural network (BPNN) for information updates was proposed.
• The VSP problem was formulated considering off-road vehicle motion characteristics and variable relationships.
• Performance was evaluated against analytical, BPNN, and recurrent neural network (RNN) methods using mining truck and loader datasets.

Main Results:

• The proposed BPNN-LSTM method demonstrated superior speed prediction accuracy compared to existing methods.
• Average prediction errors were reduced by 48.14% (vs. analytical), 35.82% (vs. BPNN), and 30.09% (vs. RNN).
• The method maintained real-time prediction performance while enhancing accuracy.

Conclusions:

• The developed double-layer BPNN-LSTM VSP method offers a significant advancement for off-road vehicle applications.
• This approach provides a new, effective solution for improving off-road vehicle speed prediction accuracy and energy management.