Closed-loop Planning on nuPlan random 14 (test)
94.8NRDiffusion-Planner
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Diffusion-PlannerType=Learning, Post-processing=True2026.03 | 94.8 | 91.7 | — | — | — | — | — | |
| CarPLANType=Learning, Post-processing=True2026.03 | 94.6 | 92.5 | — | — | — | — | — | |
| CarPlannerType=Learning, Post-processing=True2026.03 | 94.1 | 91.1 | — | — | — | — | — | |
| Log-replayType=Expert2026.01 | 94.03 | 75.86 | — | — | — | — | — | |
| Log-replayType=Expert, Post-processing=True2026.03 | 94 | 75.9 | — | — | — | — | — | |
| PLUTOType=Learning, Post-processing=True2026.03 | 92.2 | 90.2 | — | — | — | — | — | |
| SAH-DriveType=Hybrid2026.01 | 91.18 | 89.27 | — | — | — | — | — | |
| RaScType=Learning, Post-processing=True2026.03 | 91 | 82 | — | — | — | — | — | |
| PlannerRFTType=Learning2026.01 | 90.76 | 85.8 | — | — | — | — | — | |
| PDM-ClosedType=Rule, Post-processing=True2026.03 | 90.2 | 91.2 | — | — | — | — | — | |
| PDM-HybridType=Learning, Post-processing=True2026.03 | 90.2 | 91.2 | — | — | — | — | — | |
| PDM-ClosedType=Rule2026.01 | 90.05 | 91.63 | — | — | — | — | — | |
| PLUTOType=Learning2026.01 | 89.9 | 78.62 | — | — | — | — | — | |
| Flow PlannerType=Learning2026.01 | 89.88 | 82.93 | — | — | — | — | — | |
| Diffusion PlannerDPMType=Learning, Sampler=10-step DPM-solver2026.01 | 89.19 | 82.93 | — | — | — | — | — | |
| Diffusion PlannerDDIMType=Learning, Sampler=5-step DDIM2026.01 | 89.14 | 82.63 | — | — | — | — | — | |
| Diffusion-esType=Learning-based2026.01 | 88.2 | 84.2 | — | — | — | — | — | |
| ReflexDiffusionType=Learning-based2026.01 | 86.4 | 71.57 | — | — | — | — | — | |
| PlanTFType=Learning2026.01 | 85.62 | 79.58 | — | — | — | — | — | |
| GameFormerType=Learning, Post-processing=True2026.03 | 83.9 | 82.1 | — | — | — | — | — | |
| GameformerType=Hybrid2026.01 | 82.6 | 79.49 | — | — | — | — | — | |
| PlutoType=Learning-based2026.01 | 81.67 | 75.95 | — | — | — | — | — | |
| Diffusion PlannerType=Learning-based, setting=conditional dropout+cfg2026.01 | 76.44 | 70.21 | — | — | — | — | — | |
| PDM-HybridType=Hybrid2026.01 | 75.97 | 82.17 | — | — | — | — | — | |
| PDM-ClosedType=Rule-based2026.01 | 75.69 | 82.17 | — | — | — | — | — | |
| PlanTFType=Learning-based2026.01 | 73.66 | 67.81 | — | — | — | — | — | |
| Diffusion PlannerType=Learning-based2026.01 | 71.6 | 82.88 | — | — | — | — | — | |
| IDMType=Rule, Post-processing=True2026.03 | 70.4 | 74.4 | — | — | — | — | — | |
| IDMType=Rule2026.01 | 70.39 | 74.42 | — | — | — | — | — | |
| IDMType=Rule-based2026.01 | 67.61 | 64.66 | — | — | — | — | — | |
| PDM-OpenType=Learning2026.01 | 52.81 | 57.23 | — | — | — | — | — | |
| PlanCNNType=Learning-based2026.01 | 48.2 | 51.38 | — | — | — | — | — | |
| Diffusion PlannerType=Learning-based, setting=conditional dropout2026.01 | 38.45 | 44.88 | — | — | — | — | — | |
| UrbanDriverType=Learning-based2026.01 | 29.17 | 33.27 | — | — | — | — | — | |
| GameFormerType=Learning2026.01 | 11.36 | 9.31 | — | — | — | — | — | |
| CALMM-Drive-LType=Knowledge-Driven, backbone=GPT-4O, quoted=true2026.05 | — | — | 86.7 | 95.02 | 87.11 | 95.79 | — | |
| CALMM-Drive-LType=Knowledge-Driven, Base Model=GPT-4o2026.06 | — | — | 86.7 | 95.02 | 87.11 | 95.79 | — | |
| CALMM-Drive-MType=Knowledge-Driven, backbone=GPT-4O-MINI, quoted=true2026.05 | — | — | 81.82 | 92.72 | 83.37 | 95.02 | — | |
| CALMM-Drive-MType=Knowledge-Driven, Base Model=GPT-4o-mini2026.06 | — | — | 81.82 | 92.72 | 83.37 | 95.02 | — | |
| DFPTraining Paradigm=Generative IL, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 90.69 | |
| DFPTraining Paradigm=Generative IL, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 81.96 | |
| DFP-FMTraining Paradigm=Generative IL, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 90.62 | |
| DFP-FMTraining Paradigm=Generative IL, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 83.59 | |
| Diffusion PlannerTraining Paradigm=Generative IL, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 89.19 | |
| Diffusion PlannerTraining Paradigm=Generative IL, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 82.93 | |
| Diffusion-ESType=Hybrid2026.05 | — | — | 87.7 | 94.25 | 87.18 | 93.87 | — | |
| Diffusion-ESType=Hybrid2026.06 | — | — | 87.7 | 94.25 | 87.18 | 93.87 | — | |
| Diffusion-PlannerType=Data-Driven2026.05 | — | — | 88.98 | 93.87 | 83.21 | 90.42 | — | |
| Diffusion-PlannerType=Data-Driven2026.06 | — | — | 88.98 | 93.87 | 83.21 | 90.42 | — | |
| ExpertTraining Paradigm=Human log replay, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 94.03 | |
| ExpertTraining Paradigm=Human log replay, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 75.86 | |
| Flow PlannerTraining Paradigm=Generative IL, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 89.88 | |
| Flow PlannerTraining Paradigm=Generative IL, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 82.93 | |
| GameFormerType=Hybrid2026.05 | — | — | 76.89 | 85.29 | 77.8 | 86.59 | — | |
| GameFormerType=Hybrid2026.06 | — | — | 76.89 | 85.29 | 77.8 | 86.59 | — | |
| GC-PGPType=Data-Driven2026.06 | — | — | 61.95 | 72.03 | 56.46 | 65.13 | — | |
| IDMType=Rule-Based2026.05 | — | — | 70.39 | 77.01 | 72.42 | 78.54 | — | |
| IDMType=Rule-Based2026.06 | — | — | 70.39 | 77.01 | 72.42 | 78.54 | — | |
| Log-ReplayType=Expert2026.05 | — | — | 94.03 | 96.93 | 75.86 | 78.16 | — | |
| Log-ReplayType=Expert2026.06 | — | — | 94.03 | 96.93 | 75.86 | 78.16 | — | |
| LUNA-ADType=Knowledge-Driven2026.06 | — | — | 86.98 | 96.93 | 87.02 | 96.55 | — | |
| Our ApproachType=Knowledge-Driven2026.05 | — | — | 86.41 | 96.93 | 86.37 | 96.17 | — | |
| PDM-ClosedType=Rule-Based, emergency_braking=true2026.05 | — | — | 90.05 | 94.64 | 91.64 | 96.55 | — | |
| PDM-ClosedType=Rule-Based, rule-based modules for emergency braking=true2026.06 | — | — | 90.05 | 94.64 | 91.64 | 96.55 | — | |
| PDM-HybridType=Hybrid, emergency_braking=true2026.05 | — | — | 90.1 | 95.02 | 91.29 | 96.17 | — | |
| PDM-HybridType=Hybrid, rule-based modules for emergency braking=true2026.06 | — | — | 90.1 | 95.02 | 91.29 | 96.17 | — | |
| PDM-OpenTraining Paradigm=Imitation learning, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 52.81 | |
| PDM-OpenTraining Paradigm=Imitation learning, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 57.23 | |
| Plan-R1Training Paradigm=IL + RL alignment, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 91.23 | |
| Plan-R1Training Paradigm=IL + RL alignment, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 90.04 | |
| PlanTFType=Data-Driven2026.05 | — | — | 85.6 | 90.8 | 78.86 | 85.44 | — | |
| PlanTFType=Data-Driven2026.06 | — | — | 85.6 | 90.8 | 78.86 | 85.44 | — | |
| PlanTFTraining Paradigm=Imitation learning, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 85.62 | |
| PlanTFTraining Paradigm=Imitation learning, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 79.58 | |
| PLUTOType=Hybrid, emergency_braking=true2026.05 | — | — | 91.93 | 95.79 | 89.94 | 96.17 | — | |
| PLUTOType=Hybrid, rule-based modules for emergency braking=true2026.06 | — | — | 91.93 | 95.79 | 89.94 | 96.17 | — | |
| PLUTOTraining Paradigm=Imitation learning, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 89.9 | |
| PLUTOTraining Paradigm=Imitation learning, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 78.62 | |
| R2LPL-baseTraining Paradigm=Imitation learning, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 70.74 | |
| R2LPL-baseTraining Paradigm=Imitation learning, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 72.96 | |
| R2LPL-ROCL-10-bestTraining Paradigm=IL + R²LPL (envelope), Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 93.94 | |
| R2LPL-ROCL-10-bestTraining Paradigm=IL + R²LPL (envelope), Simulation Protocol=R2026.06 | — | — | — | — | — | — | 88.2 | |
| R2LPL-ROCL-5Training Paradigm=IL + R²LPL, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 92.25 | |
| R2LPL-ROCL-5Training Paradigm=IL + R²LPL, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 87.99 | |
| RasterModelType=Data-Driven2026.05 | — | — | 67.7 | 73.56 | 68.64 | 74.33 | — | |
| RasterModelType=Data-Driven2026.06 | — | — | 67.7 | 73.56 | 68.64 | 74.33 | — | |
| UrbanDriverTraining Paradigm=Imitation learning, Simulation Protocol=NR2026.06 | — | — | — | — | — | — | 51.83 | |
| UrbanDriverTraining Paradigm=Imitation learning, Simulation Protocol=R2026.06 | — | — | — | — | — | — | 67.15 | |
| UrbanDriverOLType=Data-Driven2026.05 | — | — | 63.57 | 70.88 | 60.92 | 66.28 | — | |
| UrbanDriverOLType=Data-Driven2026.06 | — | — | 63.57 | 70.88 | 60.92 | 66.28 | — |