Real-Time Vehicle Trajectory Estimation Based on Lane Change Detection using Smartphone Sensors

As technology is moving rapidly toward automation and connectivity, it is of paramount importance to predict vehicle trajectories ahead of time. This not only enhances safety but also ensures mobility in a connected and automated environment. Previous studies have shown that, given the previous trajectory, the future trajectory can be estimated. But this method suffers from considerable drawbacks in the case of intersections as it cannot predict turning movements. It also requires advanced sensors that are not readily available in most vehicles. A smartphone device can also be used in such scenarios, bringing partial automation to vehicles without these sensors. This paper presents an integrated method of estimating vehicle trajectories for both general roadway segments and intersections by using a smartphone. A lane change detection system is taken as an indicator of intersection turning movement estimation and corresponding vehicle trajectories are estimated accordingly. The system can achieve high penetration rates and can be used to replicate onboard units. Sensor readings are taken periodically which are first filtered with a low-pass filter to zero out any high-frequency noise and then fed into a machine learning model to detect lane changes. The model can successfully capture lane changes with smartphone data with high accuracy (95%). Finally, vehicle trajectory is estimated using Chebyshev's polynomial. This type of estimation system can find applications in collision prediction at intersections between a turning vehicle and a pedestrian on a crosswalk.

Automated summary

This study proposes a method to estimate vehicle trajectories using smartphones to enhance traffic safety and efficiency. The system accurately captures lane changes with smartphone data and estimates vehicle trajectories with high accuracy.