Trajectory Forecasting on nuScenes v1.0 (test)
0.897minADEkPCM
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PCMZero-shot=true, Trained on=nuPlan2026.02 | 0.897 | 2.205 | 39.8 | — | 2.857 | |
| WayformerZero-shot=true, Trained on=nuPlan2026.02 | 1.036 | 2.621 | 46.9 | — | 3.299 | |
| Forecast-MAEZero-shot=true, Trained on=nuPlan2026.02 | 1.159 | 2.673 | 60.2 | — | 3.974 | |
| G2LTrajZero-shot=true, Trained on=nuPlan2026.02 | 1.188 | 2.754 | 46.7 | — | 3.419 | |
| AutoBotZero-shot=true, Trained on=nuPlan2026.02 | 1.197 | 2.782 | 46.5 | — | 3.43 | |
| MTRZero-shot=true, Trained on=nuPlan2026.02 | 1.36 | 3.236 | 60 | — | 3.877 | |
| Li-ViP3DForecasting horizon (ΔT)=6 seconds, Candidate future trajectories (K)=6, Trajectory decoder type=regression-based2026.01 | 1.45 | 2.2 | 23.6 | 25 | — | |
| RMPZero-shot=true, Trained on=nuPlan2026.02 | 1.485 | 3.46 | 65.3 | — | 4.032 | |
| ViP3DForecasting horizon (ΔT)=6 seconds, Candidate future trajectories (K)=6, Trajectory decoder type=regression-based2026.01 | 1.52 | 2.35 | 23.9 | 23.6 | — | |
| Li-ViP3D++Forecasting horizon (ΔT)=6 seconds, Candidate future trajectories (K)=6, Trajectory decoder type=regression-based2026.01 | 1.57 | 2.53 | 26.4 | 33.5 | — | |
| SMARTZero-shot=true, Trained on=nuPlan2026.02 | 1.812 | 3.587 | 69.5 | — | 4.513 | |
| PnPNet-visionForecasting horizon (ΔT)=6 seconds, Candidate future trajectories (K)=6, Trajectory decoder type=regression-based2026.01 | 2.04 | 3.08 | 27.7 | 19.8 | — | |
| Det3D + Kalman FilterForecasting horizon (ΔT)=6 seconds, Candidate future trajectories (K)=6, Trajectory decoder type=regression-based2026.01 | 2.07 | 3.1 | 28.9 | 19.1 | — | |
| APEZero-shot=true, Trained on=nuPlan2026.02 | 3.655 | 9.214 | 88.9 | — | 9.214 |