Local Linear Wavelet Neural Network-Based Unscented Kalman Filter for Vehicle Collision Estimate Warning System and Ensuring Stable Vehicle-to-Infrastructure Communication

The accident mortality rates are rapidly increasing due to driver inattention, and traffic accidents become a significant problem on a global scale. For this reason, advanced driver assistance systems (ADASs) are essential to enhance traffic safety measures. However, adverse environmental factors, weather, and light radiation affect the sensors' accuracy. Furthermore, potential risks may go unreported if they obstruct the sensor's line of sight or are outside its limited field of view. To overcome these problems, this research presents a vehicle collision estimate warning system that leverages a combined approach of a local linear wavelet neural network (LLWNN) and an unscented Kalman filter (UKF). The system integrates sensor data and vehicle-to-everything (V2X) communication to enhance the accuracy and reliability of vehicle state estimation and collision prediction. The LLWNN module is responsible for forecasting the future states of the vehicle based on historical sensor measurements. This powerful time-series modeling technique allows the system to anticipate the vehicle's trajectory and potential collision risks. The UKF then optimally fuses the LLWNN predictions with the real-time sensor data, including information received through V2X communication, to generate accurate, up-to-date estimates of the vehicle's state. The V2X technology enables the seamless exchange of critical safety information between the host vehicle, surrounding vehicles, infrastructure, and other road users. This includes data on vehicle position, speed, acceleration, and intended maneuvers. By incorporating this V2X-based situational awareness, the system can better perceive the dynamic traffic environment and identify potential collision threats that may be outside the line of sight or detection range of the vehicle's onboard sensors alone. The LLWNN-based UKF module then processes this rich, multimodal data to provide timely and pertinent collision alerts to the driver. These alerts can warn the driver of impending collisions with distant objects, enabling them to take appropriate evasive action. By implementing this integrated LLWNN-UKF approach leveraging sensor data and V2X communication, we aim to reduce the number of collisions caused by reckless driving, which will lead to a decrease in traffic-related fatalities and injuries.