Vehicle Bridge Wireless Load Monitoring System Based on Polynomial Regression

Overloading of vehicles is a common phenomenon in society, exhibiting high concealment during normal operation and thus being difficult to detect. Therefore, this paper proposes a novel load collection and wireless communication system that integrates wireless communication and machine learning to achieve real-time load prediction for small freight vehicles. The system is based on radio frequency (RF) communication. It collects axle stress data from the freight vehicle, transmits the acquired data wirelessly to a receiver, and then forwards it to the on-board computer via Controller Area Network (CAN) communication. Ultimately, a machine learning model is employed to predict the vehicle load, thereby enabling load monitoring. The proposed method is based on a lightweight algorithmic model. Considering the requirements for long-term operation, a low-power design is implemented to minimize deployment costs while ensuring fundamental computational capability. By employing a polynomial regression prediction algorithm and evaluating its overall performance on a real-world load dataset, the optimal data collection points are identified. Experimental validation demonstrates that the mean squared error of the prediction results is 0.3149, and the cross-validation score reaches 0.9233. Furthermore, the maximum error within the 95% confidence interval accounts for only 1.658% of the maximum axle load. These results confirm the system's capability to accurately predict actual load data.