LSTM-Based Virtual Load Sensor for Heavy-Duty Vehicles.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Automotive Engineering

Background:

• Accurate estimation of heavy vehicle mass is crucial for logistics and safety.
• Traditional load sensors can be expensive and complex to install.
• Developing virtual sensing technologies offers a cost-effective alternative.

Purpose of the Study:

• To design and evaluate a virtual load sensor for estimating heavy vehicle mass.
• To utilize a Long Short-Term Memory (LSTM) recurrent neural network for mass estimation.
• To assess the accuracy of the proposed LSTM-based method against commercial sensors.

Main Methods:

• A two-layer Long Short-Term Memory (LSTM) network was developed.
• The LSTM network utilized vehicle speed, acceleration, engine speed, torque, and pedal position as inputs.
• The model was trained using data from the Truckmaker simulation environment and tested on Worldwide harmonized Light vehicles Test Cycle (WLTC) data.

Main Results:

• The LSTM-based virtual load sensor accurately estimated heavy vehicle mass.
• The estimation achieved accuracy comparable to commercial load sensors.
• The system demonstrated reliable performance across a four-ton range of load masses.

Conclusions:

• A Long Short-Term Memory (LSTM) network can effectively serve as a virtual load sensor for heavy vehicles.
• The proposed method provides a promising, accurate, and potentially more economical approach to vehicle mass estimation.
• This technology has significant implications for intelligent transportation systems and fleet management.