From Proxy Learning to Driving Decisions: A Transfer-Based Framework for Evaluating Future-Aware Autonomous Driving Planners
推荐理由
Future-aware representations and world models are increasingly used in proposal-based autonomous-driving planners to improve trajectory selection. However, imp…
核心判断
论文摘要（中文）
论文提出 Proxy-to-Decision Transfer（PDT）分析框架，用于检验世界模型等未来信息是否真的转化为自动驾驶规划收益。框架分解决策转移中的价值损失，并设置配对、最小效应、扩大样本、安全非补偿和跨配置稳健性等验证条件。
Future-aware representations and world models are increasingly used in proposal-based autonomous-driving planners to improve trajectory selection. However, improvements in proxy objectives or restricted subsets are often interpreted as planning gains without verifying proposal ordering, selected trajectories, full-scale utility, and critical driving components. We propose the Proxy-to-Decision Transfer (PDT) Framework, an analysis framework that evaluates when learned future information supports a reliable driving-performance improvement claim. Its Decision-Transfer Decomposition Module localizes value loss through score margins, switch-conditioned utility, and support-versus-selection regret. Its ReliabilityConstrained Validation Module requires exact pairing, a minimum meaningful effect, scaleexpanded confirmation, safety non-compensation, sequential comparability, and family-level robustness. On a representative future-aware planner evaluated with NAVSIM-v1, component BCE decreases from 0.705 to 0.530 while held selected PDM decreases from 0.963 to 0.961. A separate candidate improves a 512-record prefix by 0.00909, with a scene-bootstrap 95% interval of [0.000744, 0.0177], but its 2048-record and complete-support intervals include zero. A proposal-level replay further confirms the switch-utility decomposition, yet none of 432 screened configurations passes the two-half, two-seed robustness gate. PDT therefore identifies where decision transfer fails or remains indeterminate across proxy, subset, aggregate, and selection evidence.
研究动机（中文总结）
规划论文常把代理目标改善或小规模子集上的提升直接解释成驾驶规划变好，却没有验证候选排序、最终轨迹选择、全量效用和安全关键指标是否同步改善。
We propose the Proxy-to-Decision Transfer (PDT) Framework, an analysis framework that evaluates when learned future information supports a reliable driving-performance improvement claim.
创新与贡献（中文总结）
将“未来表征是否有用”拆成可定位的决策转移环节，并用可靠性约束验证，区分统计符号、实际效应、组件不补偿和跨配置稳健性。
We propose the Proxy-to-Decision Transfer (PDT) Framework, an analysis framework that evaluates when learned future information supports a reliable driving-performance improvement claim.
方法与证据
方法（中文总结）
决策转移分解模块分析分数间隔、选择切换条件下的效用和支持集/选择集 regret；可靠性约束验证模块检查严格配对、最小有意义效应、扩大样本确认、安全指标、顺序可比性及家族稳健性。
We propose the Proxy-to-Decision Transfer (PDT) Framework, an analysis framework that evaluates when learned future information supports a reliable driving-performance improvement claim.
实验结果（中文总结）
在 NAVSIM-v1 案例中，代理 BCE 从 0.705 降至 0.530，但 held-selected PDM 从 0.963 变为 0.961；432 种筛选配置均未通过双半样本、双随机种子的稳健性门槛，因此论文结论是尚无稳健提升证据，而非证明方法提升规划性能。
Finally, the aggregate-score and calibration experiments show that a machine-precision sign, a practically meaningful effect, component non-compensation, and family-level robustness are separate objectives.