Lane-change prediction remains a critical challenge for automated driving systems. Traditional machine learning approaches learn statistical associations between observed driving variables and future maneuvers but often overlook causal dependencies among input variables. This limits interpretability, especially when physically related variables such as longitudinal gap, relative longitudinal velocity, and Time-To-Collision (TTC) are treated as independent flat inputs.
To address this, a team of researchers — including Manzour, Mohamed, Kumar Aditya, Augusto Luis Ballardini, and Miguel Ángel Sotelo — has proposed a causal-inference-based framework for lane-change prediction and explanation. The work, published as a preprint on arXiv, combines linguistic feature construction, expert-constrained causal discovery, deep structural causal modeling with Deep End-to-end Causal Inference (DECI), intervention-based effect analysis, refutation testing, and recursive causal-chain explanation.
Prediction Performance and Beyond
The framework achieves average F1-scores above 95% during the first three seconds before the lane-marking crossing event. However, the objective extends beyond accuracy. It aims to identify candidate variables that directly contribute to the prediction, the upstream factors influencing them, and the causal chains through which effects propagate.
Causal Reasoning Components
The framework uses intervention-based effect analysis to distinguish influential from weakly influential variables under the learned causal structure. It further distinguishes candidate direct contributors from mediated effects and generates contrastive causal-chain explanations. These explanations clarify why the predicted maneuver is favored and why alternative maneuvers are less supported.
Implications for Automated Driving
The main contribution is a mechanism-aware lane-change prediction pipeline that moves beyond correlation-based classification toward more interpretable causal reasoning for maneuver prediction. This could enable safer decision-making in intelligent vehicles by providing transparent insights into prediction logic.
| Key Metric | Value |
|---|---|
| Average F1-score | Above 95% |
| Prediction window | First 3 seconds before lane-marking crossing |
| Methodology | Causal discovery, DECI, intervention analysis, refutation testing |
| Goal | Distinguish direct vs. mediated effects, generate causal-chain explanations |
The research underscores the importance of causal inference in autonomous vehicle systems, offering a path toward more reliable and interpretable predictions.
