Yao Long, Teng (Nanyang Technological University), Naing, Htet (Nanyang Technological University), Cai, Wentong (NANYANG TECHNOLOGICAL UNIVERSITY)

Integrating Data and Rules: A Hybrid Approach for Robust Lane Change Intention Prediction under Distribution Shift

Scheduled for presentation during the Poster Session "Detection, estimatation and prediction for intelligent transportation systems" (WeAT17), Wednesday, September 25, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Lane change intention prediction is critical in improving road safety. In general, this task is achieved via two distinct approaches: The data-based approach and the rule-based approach. Generally, the former outperforms the latter under in-distribution scenarios. However, under out-distribution scenarios, data-based methods tend to perform extremely poorly. Rule-based methods are robust to such problems since they are formulated based on theory. Instead of relying exclusively on each approach, it is more advantageous to combine them to exploit both of their strengths. Thus, in this paper a physics guided hybrid model is proposed for lane change intention prediction to improve model robustness under distribution shift. The effectiveness of the proposed model is shown in the experimental results where the proposed model outperforms rule-based models under in-distribution scenarios by 4%, and outperforms both rule-based models and databased models in out-distribution scenarios by 3% and 60% respectively in the highD dataset.