Vehicle Trajectory Prediction Algorithm Based on Hybrid Prediction Model with Multiple Influencing Factors

Area of Science:

• Autonomous Driving Systems
• Artificial Intelligence in Transportation
• Robotics and Control

Background:

• Vehicle trajectory prediction is crucial for safe autonomous driving.
• Existing methods often overlook interactions between surrounding vehicles.
• This limitation impacts the dynamic perception and cooperative capabilities of autonomous systems.

Purpose of the Study:

• To propose a novel hybrid prediction algorithm for vehicle trajectory.
• To enhance autonomous driving safety by improving prediction accuracy.
• To address the limitations of current models in capturing vehicle interactions.

Main Methods:

• Utilized a two-layer long short-term memory (LSTM) network for context extraction.
• Developed a fusion module to integrate temporal, spatial, and interactive influences.
• Implemented a prediction module to output coordinate-based movement positions.

Main Results:

• The proposed hybrid model demonstrated superior prediction accuracy.
• Validation on I-80 and US-101 datasets confirmed algorithm's effectiveness.
• The model successfully captured and utilized surrounding vehicle interactions for improved predictions.

Conclusions:

• The hybrid trajectory prediction algorithm offers enhanced precision for autonomous vehicles.
• Considering multi-faceted influences (temporal, spatial, interactive) is key to accurate predictions.
• This approach advances the development of safer and more capable autonomous driving systems.