Fuzzy Logic-Based Probabilistic Forecasting of Short-Term Trajectory for Autonomous Vehicles

The rapid advancement of Intelligent Transportation Systems (ITS) has led to the widespread adoption of autonomous vehicles as common consumer electronics. Accurate and real-time vehicle trajectory forecasting significantly enhances the potential of these vehicles. However, traditional time series forecasting methods are susceptible to noise and require linear sequences, making them inadequate for integrating traffic rules and environmental factors. This limitation often results in unreliable multi-modal predictions and reduced prediction accuracy. To address these challenges, this paper introduces a Probabilistic Trajectory Forecasting approach (PTFO), specifically designed for short-term trajectory forecasting. By analyzing actual vehicle trajectory data, we applied fuzzy logic to demonstrate that vehicle trajectories follow a Gaussian distribution in the short term. Based on this finding, we further categorize vehicle trajectory motion into two distinct modes: single mode and complex mode, to better capture the variability in vehicular motion. Additionally, we conduct a thorough investigation into how vehicular spatial relationships and extrinsic factors influence trajectory predictions, thereby improving prediction accuracy. The PTFO algorithm’s strengths lie in its ability to incorporate various motion modes and hidden factors affecting future trajectories. Simulation results indicate that the PTFO algorithm outperforms other probabilistic trajectory forecasting techniques.