Research KB 登录

EVI-026 Evidence

Tables 2–3:interPlan 零样本 G2DP† 62.55(+9.65),collision +10.15、TTC +10.75;progress/speed limit 略降。

id
updated
type evidence
claims CLM-026
source PPR-2606.26017
observation Table 2 (tab:interplan_results) overall + interactive subsets; Table 3 (tab:grid_guidance_benchmark) nuPlan-metric breakdown vs Diffusion Planner
method interPlan 335 handcrafted interactive/adversarial scenarios; nuPlan-trained model zero-shot, no tuning; nuPlan metrics recomputed on interPlan
scope zero-shot transfer; interactive subsets Nudge Around / High Traffic / Jaywalk

Result

Overall:Diffusion Planner 52.90、G2DP 61.74(+8.84)、G2DP\({}^\dagger\) 62.55(+9.65)。分项(G2DP\({}^\dagger\) vs DP):Collision avoidance 94.93(+10.15)、TTC 94.33(+10.75)、Drivable area 93.73(+3.88)、Comfort 58.21(+4.78);Progress 94.43(−0.20)、Speed limit 98.21(−0.31)、Driving direction G2DP 97.91(−1.19)。子集:Jaywalk G2DP 51.12 最高(Flow Planner 43.57);Nudge Around Flow Planner 72.96 最高。

Caveats

单种子、无 tuning 复试;Nudge Around 输给 Flow Planner 说明优势依赖场景类型;interPlan 用 nuPlan 指标复算,分项可比性以该协议为前提。

关联(12)

  • PPR-2404.07569 Can Vehicle Motion Planning Generalize to Realistic Long-tail Scenarios?(interPlan)
  • PPR-2501.15564 Diffusion-Based Planning for Autonomous Driving with Flexible Guidance
  • PPR-2606.26017 G2DP: Diffusion Planning with Spatio-Temporal Grid Guidance
  • FRM-2606.26017 G2DP 五层重建:核心对象是 CPT-013 时空代价体(占据概率栅格 × 路线进度图融合)在推理期的可微引导;公理层为能量重加权目标分布、Top-K 车体聚合与末段注入等六个作者前提;论证主干是 Table 4 在同一 DiT 骨干下隔离『稠密网格 vs 稀疏实例』的引导模态消融(稀疏臂按 PPR-2501.15564 附录 C 实现、以 oracle 邻车供给)。
  • CLM-026 nuPlan 训练的模型零样本迁移 interPlan:相对共享生成管线的 Diffusion Planner 大幅领先,增益集中于安全维度(碰撞避免、TTC),progress/speed-limit 类项持平或微降;优势依场景类型非全域均匀。
  • EVI-024 co-source Table 1:G2DP 系居纯 IL planner 首位,Test14-hard R 77.61(+7.2 over Flow Planner,+8.4 over Diffusion Planner)。
  • EVI-025 co-source Table 4:Grid +3.35 score/+3.59 collision/+4.94 TTC;Sparse(Oracle) −1.96 score/−10.13 comfort/+0.64 collision。稀疏臂按 PPR-2501.15564 附录 C 式 10 的实例级有向距离能量实现,与稠密臂在信号形态与信息内容两轴同时不同(评注已补)。
  • EVI-027 co-source Table 5:G2DP 较 DP collision +11.0、progress +10.5、TTC +6.7;progress/along-route 超 PDM-Closed。
  • EVI-028 co-source Fig. 8:steps 8–9 + lambda_t=0.5 峰值 77.61;早窗与 lambda_t≥0.8 均退化。
  • EVI-029 co-source Table 6:Top-15 双split峰值 92.26/80.05;Average 86.33 洗信号、Max 76.84 抗伪影差。
  • EVI-030 co-source Fig. 7:γ=0.95 峰值;γ=1.0 过度保守刹车致高密度区追尾,γ=0.2 安全弱化。
  • EVI-031 co-source Table 1/2 + 正文:G2DP† reactive +3.7、interPlan +0.81、精修依赖 +4.1 vs DP +12.8。