🚗 End-to-End Driving with Online Trajectory Evaluation via BEV World Model (ICCV2025)

Yingyan Li*, Yuqi Wang*, Yang Liu, Jiawei He, Lue Fan† and Zhaoxiang Zhang† [📄 Paper (arXiv:2504.01941)]

We present WoTE, an end-to-end autonomous driving framework that leverages a BEV-based world model to predict future agent states, enabling online trajectory evaluation and selection.

---

📊 1. Results & Checkpoints

Method	NC	DAC	EP	TTC	Comfort	PDMS	Training Time	GPU Memory	Checkpoint
WoTE	98.5	96.8	81.9	94.9	99.9	88.3	3 hrs	22 GB	📥 Download

Training conducted on 8 NVIDIA L20 GPUs.

Legend
• NC: No Collision
• DAC: Drivable Area Compliance
• EP: Ego Progress
• TTC: Time to Collision
• Comfort: Comfort • PDMS: Predictive Driver Model Score

---

📦 2. Dataset & File Structure

root/
├── ckpts/
│   └── resnet34.pth
├── dataset/
│   ├── maps/
│   ├── navsim_logs/
│   │   ├── test/
│   │   └── trainval/
│   ├── sensor_blobs/
│   │   ├── test/
│   │   └── trainval/
│   └── extra_data/planning_vb/
│       ├── trajectory_anchors_256.npy
│       └── formatted_pdm_score_256.npy
└── exp/
    └── metric_cache/

📁 a. Download NAVSIM Dataset

To obtain the navsim dataset:

bash download/download_maps.sh
bash download/download_navtrain.sh
bash download/download_test.sh

📁 b. Prepare Extra Data

You can either download the extra data directly from 🌐 Google Drive or generate the files manually:

• K-means anchor trajectories (trajectory_anchors_256.npy):

python scripts/misc/k_means_trajs.py

• Reward generation (formatted_pdm_score_256.npy):

bash scripts/misc/gen_pdm_score.sh

📁 c. Precompute Metric Cache

bash scripts/evaluation/run_metric_caching.sh

---

⚙️ 3. Installation

Create the conda environment:

conda env create -f environment.yml
conda activate wote

Install dependencies:

pip install -r requirements.txt
pip install git+https://github.com/motional/nuplan-devkit.git@nuplan-devkit-v1.2#egg=nuplan-devkit

Add environment variables to ~/.bashrc (modify paths as needed):

export NUPLAN_MAP_VERSION="nuplan-maps-v1.0"
export NUPLAN_MAPS_ROOT="$HOME/navsim_workspace/dataset/maps"
export NAVSIM_EXP_ROOT="$HOME/navsim_workspace/exp"
export NAVSIM_DEVKIT_ROOT="$HOME/navsim_workspace/"
export OPENSCENE_DATA_ROOT="$HOME/navsim_workspace/dataset"

---

🚀 4. Training & Evaluation

Update paths in:

navsim/agents/WoTE/configs/default.py

Then launch training:

bash scripts/training/run_wote.sh

Evaluation:

bash scripts/evaluation/eval_wote.sh

---

---

🔍 5. More from Us

If you're interested in world models for autonomous driving, feel free to check out our related work:

• LAW (ICLR 2025): Using latent world models for self-supervised feature learning in end-to-end autonomous driving.