Temel Kavramlar
This study proposes an innovative trajectory prediction model that integrates cognitive insights on perceived safety and dynamic decision-making to enhance the performance of autonomous driving systems in mixed autonomy environments.
Özet
This paper introduces a novel trajectory prediction model for autonomous driving that infuses cognitive insights on perceived safety and dynamic decision-making. The model consists of three key modules:
-
Perceived Safety-Aware Module:
- Quantitative Safety Assessment (QSA): Evaluates the safety of a driving scenario using metrics like Time-to-Collision (TTC), Time Exposed TTC (TET), and Time Integrated TTC (TIT).
- Driver Behavior Profiling (DBP): Captures the driving behavior of individual agents in real-time using dynamic graph-based centrality measures, without the need for manual labeling.
-
Priority-Aware Module:
- Encodes the spatial positioning of vehicles and their influence on the target vehicle's trajectory.
-
Interaction-Aware Module:
- Utilizes a lightweight transformer-based framework, Leanformer, to capture the intricate inter-vehicular interactions.
The proposed model demonstrates significant performance improvements over state-of-the-art baselines on several key datasets, including NGSIM, HighD, and MoCAD. It achieves gains of 16.2%, 27.4%, and 19.8%, respectively, in long-term trajectory prediction. Moreover, the model exhibits exceptional robustness in handling incomplete or inconsistent data, outperforming most baselines in such scenarios. This adaptability and resilience position the proposed model as a viable tool for real-world autonomous driving systems, advancing the state-of-the-art in vehicle trajectory prediction for enhanced safety and efficiency.
İstatistikler
Time-to-Collision (TTC) is a measure used to evaluate the time available before two vehicles collide if they continue on their current trajectories.
Time Exposed Time-to-Collision (TET) measures the exposure duration to critical TTC values within a given time horizon.
Time Integrated Time-to-Collision (TIT) integrates the TTC profile to evaluate safety levels, factoring in the evolution of each vehicle's TET temporally.
Risk Tendency Index (RTI) captures the subjective risk perception and dynamic risk volatility between vehicles.
Alıntılar
"Is the key to advancing AD not just in accumulating more data or refining algorithms, but in gaining a deeper understanding of the driving environment itself? How can we reshape our models to interpret and respond to the intricate human dynamics that underpin driving?"
"By doing so, we aim to enhance the models' ability to interpret driving behaviors, leading to predictions that are not only technically accurate but also richly informed by contextual and psychological insights."