GRaD-Nav Series

[IROS 2025] GRaD-Nav: Learning Visual Drone Navigation with Gaussian Radiance Fields and Differentiable Dynamics

Project Page | Video | Arxiv

---

[RA-L 2025] GRaD-Nav++: Vision-Language Model Enabled Visual Drone Navigation with Gaussian Radiance Fields and Differentiable Dynamics

Project Page | Video | Arxiv

---

Configurations

• Clone the repository:

  git clone https://github.com/Qianzhong-Chen/grad_nav.git
  

Set up conda environment and install dependencies.

• Install and configure nerfstudio, start nerfstudio conda env

  conda activate nerfstudio
  pip install gsplat
  

• Install gym and transformers

  pip install gym
  pip install transformers
  

Data Download and Setup

• Download the required data.

  cd grad_nav/envs/assets
  gdown --id 1xIDb1-HFkniBagvMbWDgfy4okKTckwra # point cloud data
  unzip point_cloud.zip
  gdown --id 1skbXqg__Ew3ytNmqe_xiaqHpexZnsfdw # 3DGS data
  unzip gs_data.zip
  cd ../../
  

Training

1. GRaD-Nav:

• GRaD-Nav long trajectory training:

  python examples/train_gradnav.py --cfg examples/cfg/gradnav/drone_long_traj.yaml --logdir examples/logs/DroneLongTraj/gradnav
  

• GRaD-Nav multi-gate training:

  python examples/train_gradnav.py --cfg examples/cfg/gradnav/drone_multi_gate.yaml --logdir examples/logs/DroneMultiGate/gradnav
  

2. GRaD-Nav++:

• GRaD-Nav VLA long task training:

  python examples/train_gradnav_vla_moe.py --cfg examples/cfg/gradnav_vla_moe/drone_long_task.yaml --logdir examples/logs/DroneVLALongTaskEnv/gradnav_MoE
  

• GRaD-Nav VLA multi map training:

  python examples/train_gradnav_vla_moe.py --cfg examples/cfg/gradnav_vla_moe/drone_multi_map.yaml --logdir examples/logs/DroneVLAMultiMapEnv/gradnav_MoE
  

3. Baselines

• PPO long trajectory training:

  python examples/train_ppo.py --cfg examples/cfg/ppo/drone_ppo.yaml --logdir examples/logs/DronePPO/ppo
  

• BPTT long trajectory training:

  python examples/train_bptt.py --cfg examples/cfg/bptt/drone_long_traj.yaml --logdir examples/logs/DroneLongTraj/bptt
  

• Using wandb
  remove os.environ["WANDB_MODE"] = "disabled" from algorithms/<algorithm_name>.py

Testing

1. GRaD-Nav:

• GRaD-Nav long trajectory testing:

  python examples/train_gradnav.py --cfg examples/cfg/gradnav/drone_long_traj.yaml --checkpoint examples/logs/DroneLongTraj/gradnav/<map_name>/<time_stamp>/best_policy.pt --play --render
  

• GRaD-Nav multi-gate testing:

  python examples/train_gradnav.py --cfg examples/cfg/gradnav/drone_multi_gate.yaml --checkpoint examples/logs/DroneMultiGate/gradnav/<map_name>/<time_stamp>/best_policy.pt --play --render
  

2. GRaD-Nav++:

• GRaD-Nav VLA long task testing single task (changing task in drone_long_task.yaml):

  python examples/train_gradnav_vla_moe.py --cfg examples/cfg/gradnav_vla_moe/drone_long_task.yaml --checkpoint examples/logs/DroneVLALongTaskEnv/gradnav_MoE/<map_name>/<time_stamp>/final_policy.pt --play --render
  

• GRaD-Nav VLA long task testing all tasks:

  python auto_test_vla_long_task_moe.py --checkpoint examples/logs/DroneVLALongTaskEnv/gradnav_MoE/<map_name>/<time_stamp>/final_policy.pt --play --render
  

• GRaD-Nav VLA multi map testing single task (changing task in drone_multi_map.yaml):

  python examples/train_gradnav_vla_moe.py --cfg examples/cfg/gradnav_vla_moe/drone_multi_map.yaml --checkpoint examples/logs/DroneVLAMultiMapEnv/gradnav_MoE/<map_name>/<time_stamp>/final_policy.pt --play --render
  

• GRaD-Nav VLA multi map testing all tasks:

  python auto_test_vla_multi_map_moe.py --checkpoint examples/logs/DroneVLAMultiMapEnv/gradnav_MoE/<map_name>/<time_stamp>/final_policy.pt --play --render
  

3. Baselines

• PPO long trajectory testing:

  python examples/train_ppo.py --cfg examples/cfg/ppo/drone_ppo.yaml --checkpoint examples/logs/DronePPO/ppo/<map_name>/<time_stamp>/best_policy.pt --play --render
  

• BPTT long trajectory testing:

  python examples/train_bptt.py --cfg examples/cfg/bptt/drone_long_traj.yaml --checkpoint examples/logs/DroneLongTraj/bptt/<map_name>/<time_stamp>/best_policy.pt --play --render
  

4. Test results

• Simulation results can be found at examples/outputs/<env_name>/<map_name>/<time_stamp>/

Citation

If you find our paper or code is useful, please consider citing:

  @misc{chen2025gradnavefficientlylearningvisual,
        title={GRaD-Nav: Efficiently Learning Visual Drone Navigation with Gaussian Radiance Fields and Differentiable Dynamics}, 
        author={Qianzhong Chen and Jiankai Sun and Naixiang Gao and JunEn Low and Timothy Chen and Mac Schwager},
        year={2025},
        eprint={2503.03984},
        archivePrefix={arXiv},
        primaryClass={cs.RO},
        url={https://arxiv.org/abs/2503.03984}, 
    }

  @article{chen2025grad,
    title={GRaD-Nav++: Vision-Language Model Enabled Visual Drone Navigation with Gaussian Radiance Fields and Differentiable Dynamics},
    author={Chen, Qianzhong and Gao, Naixiang and Huang, Suning and Low, JunEn and Chen, Timothy and Sun, Jiankai and Schwager, Mac},
    journal={arXiv preprint arXiv:2506.14009},
    year={2025}
}