RFID Data-Driven Vehicle Speed Prediction via Adaptive Extended Kalman Filter

Area of Science:

• Transportation Engineering
• Signal Processing
• Machine Learning

Background:

• Traditional vehicle speed prediction relies on GPS and video, susceptible to environmental interference.
• Environmental factors like weather and electromagnetic waves significantly impact prediction accuracy.
• There is a need for robust vehicle speed prediction methods independent of environmental conditions.

Purpose of the Study:

• To develop and evaluate a novel Radio Frequency Identification (RFID) data-driven vehicle speed prediction algorithm.
• To enhance the accuracy and reliability of vehicle speed estimation.
• To improve upon existing Kalman filter techniques for dynamic systems.

Main Methods:

• Utilized an on-board RFID reader to collect vehicle speed and time data from road-deployed tags.
• Implemented an adaptive extended Kalman filter (AEKF) algorithm for vehicle speed prediction.
• Compared the performance of AEKF against the conventional extended Kalman filter (EKF) using simulation.

Main Results:

• The AEKF algorithm demonstrated improved dynamic filtering performance compared to the conventional EKF.
• AEKF effectively suppressed filtering divergence, leading to more stable predictions.
• Significant improvements in prediction accuracy were observed: 57.4% reduction in Mean Square Error (MSE) and 32.4% in Mean Absolute Error (MAE).

Conclusions:

• The proposed AEKF algorithm offers a more accurate and reliable method for vehicle speed prediction using RFID data.
• AEKF overcomes limitations of traditional methods by reducing environmental dependency.
• This approach enhances vehicle state estimation, crucial for intelligent transportation systems.