maxim-igenbergs/tcp-carla-repro

TCP – Trajectory-Guided Control Prediction (Reproduced Checkpoint)

Reproduced checkpoint of TCP: Trajectory-Guided Control Prediction for End-to-End Autonomous Driving (Wu et al., NeurIPS 2022), used for evaluation in the bachelor's thesis: Dual-Axis Testing of Visual Robustness and Topological Generalization in Vision-based End-to-End Driving Models.

Model Description

TCP combines trajectory prediction with direct control prediction in a unified architecture. The model predicts both a future trajectory and vehicle controls, using the trajectory branch to guide the control prediction through a multi-step reasoning approach.

Checkpoint

• best_model.ckpt

Origin

Unofficial reproduction shared in TCP GitHub issue #11 (Oct 2022). Not affiliated with or endorsed by the original authors or OpenDriveLab.

Related

• DAVE-2 Model
• DAVE-2-GRU Model
• ViT Driving Model
• Evaluation Runs Dataset