Vision-Language Navigation on Various Vision-Language Navigation Datasets
40,801Number of TrajectoriesUAV-Flow
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UAV-FlowAction Space=6 DoF, Scene Environment=Real World, UnrealCV, Traj Generation=Human Control, Traj Annotation=Human Annotation + LLM, Kinematics=Yes, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 40,801 | — | |
| IndoorUAVAction Space=4 DoF, Scene Environment=Mp3D, Gibson, HM3D, Replica, Traj Generation=Human Control, Traj Annotation=LLM-based, Kinematics=No, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 34,925 | 1,075 | |
| CityNavAction Space=4 DoF, Scene Environment=SensatUrban, Traj Generation=Human Control, Traj Annotation=Pre-existed Dataset (City Refer), Kinematics=No, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 32,637 | 34 | |
| RxRAction Space=Node-based, Scene Environment=Matterport3D, Traj Generation=Sampling from Nodes, Traj Annotation=Human Annotation, Kinematics=No, Scene Extensibility=Manual Scan, Traj Extensibility=Manual Scan2026.05 | 13,992 | 90 | |
| OpenUAVAction Space=6 DoF, Scene Environment=AirSim + UE, Traj Generation=Human Control, Traj Annotation=LLM-based, Kinematics=Yes, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 12,149 | 22 | |
| AerialVLNAction Space=4 DoF, Scene Environment=AirSim + UE, Traj Generation=Human Control, Traj Annotation=Human Annotation, Kinematics=No, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 8,446 | 25 | |
| R2RAction Space=Node-based, Scene Environment=Matterport3D, Traj Generation=Sampling from Nodes, Traj Annotation=Human Annotation, Kinematics=No, Scene Extensibility=Manual Scan, Traj Extensibility=Manual Scan2026.05 | 7,189 | 90 | |
| AVDNAction Space=3 DoF, Scene Environment=xView, Traj Generation=Human Control, Traj Annotation=Human Annotation, Kinematics=No, Scene Extensibility=Manual Record, Traj Extensibility=Manual Record2026.05 | 6,269 | — | |
| VLN-CEAction Space=2 DoF, Scene Environment=Matterport3D, Traj Generation=A* between R2R Path Nodes, Traj Annotation=Adapted from R2R, Kinematics=No, Scene Extensibility=Manual Scan/Auto Record, Traj Extensibility=Manual Scan/Auto Record2026.05 | 4,475 | 90 | |
| LHPR-VLNAction Space=2 DoF, Scene Environment=HM3D, Traj Generation=D* Lite-based, Traj Annotation=LLM-based, Kinematics=No, Scene Extensibility=Manual Scan/Auto Record, Traj Extensibility=Manual Scan/Auto Record2026.05 | 3,260 | 216 | |
| SAGE-3DAction Space=2 DoF, Scene Environment=3DGS, Traj Generation=A*, Traj Annotation=LLM-based, Kinematics=No, Scene Extensibility=Manual Scan/Auto Record, Traj Extensibility=Manual Scan/Auto Record2026.05 | 100 | 1,000 | |
| OpenFlyAction Space=4 DoF, Scene Environment=AirSim, GTA5, 3DGS, GE, Traj Generation=A*, Traj Annotation=LLM-based, Kinematics=No, Scene Extensibility=Manual Scan/Record, Traj Extensibility=Manual Scan/Record2026.05 | 100 | 18 | |
| FlyMirageAction Space=6 DoF, Scene Environment=3DGS, Traj Generation=Dynamically Feasible UAV Planner, Traj Annotation=LLM-based, Kinematics=Yes, Scene Extensibility=Fully Automated, Traj Extensibility=Fully Automated2026.05 | 50 | 500 |