What question did this study set out to answer?

To develop an end-to-end multi-view lateral control framework that dynamically weights camera views based on navigational intent and road complexity.

August 16, 2026Journal of Computational Design and EngineeringOpen Access

End-to-End Vehicle Lateral Control with Transformer-Based Perception and Intention-Aware View Weighting

View Full Paper

Ask AI

Bookmark

Authors

DKDong-Hyun Kim YLYong‐Gu Lee

Discussion

Member takes

Overview

Simulation study demonstrates over 95% driving autonomy in unseen urban environments using intention-aware multi-view weighting, highlighting enhanced lateral stability for autonomous vehicles.

Key Points

To develop an end-to-end multi-view lateral control framework that dynamically weights camera views based on navigational intent and road complexity.
Fused multi-view RGB imagery with YOLOPv2 semantic segmentation across left, front, and right perspectives using a hybrid Vision Transformer embedding.
Engineered a ViewWeightGater combining high-level command attention with a lane density-based complexity prior via a Hill function, coupled with a GRU velocity encoder and Temporal Transformer.
Trained the framework on 14,873 multi-view driving sequences in CARLA Town05 and evaluated lateral control performance across unseen Town01 and Town02 environments.
Achieved over 95% autonomy in unseen urban test environments while demonstrating human-like view attention allocation.
Ablation experiments verified that semantic perception fusion and complexity-adaptive view gating provide complementary contributions to lateral driving stability.

Cite This Study

Kim et al. (2026) studied this question.

synapsesocial.com/papers/6a8179abf2fb91fc834acde8 https://doi.org/10.1093/jcde/qwag075

View Full Paper

Ask AI

Bookmark

August 16, 2026Journal of Computational Design and EngineeringOpen Access

End-to-End Vehicle Lateral Control with Transformer-Based Perception and Intention-Aware View Weighting

View Full Paper

Ask AI

Bookmark

Authors

DKDong-Hyun Kim YLYong‐Gu Lee

Discussion

Member takes

Overview

Simulation study demonstrates over 95% driving autonomy in unseen urban environments using intention-aware multi-view weighting, highlighting enhanced lateral stability for autonomous vehicles.

Key Points

To develop an end-to-end multi-view lateral control framework that dynamically weights camera views based on navigational intent and road complexity.
Fused multi-view RGB imagery with YOLOPv2 semantic segmentation across left, front, and right perspectives using a hybrid Vision Transformer embedding.
Engineered a ViewWeightGater combining high-level command attention with a lane density-based complexity prior via a Hill function, coupled with a GRU velocity encoder and Temporal Transformer.
Trained the framework on 14,873 multi-view driving sequences in CARLA Town05 and evaluated lateral control performance across unseen Town01 and Town02 environments.
Achieved over 95% autonomy in unseen urban test environments while demonstrating human-like view attention allocation.
Ablation experiments verified that semantic perception fusion and complexity-adaptive view gating provide complementary contributions to lateral driving stability.

Cite This Study

Kim et al. (2026) studied this question.

synapsesocial.com/papers/6a8179abf2fb91fc834acde8 https://doi.org/10.1093/jcde/qwag075

View Full Paper

Ask AI

Bookmark