arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2610.36285cs.RO
收藏

ASPEN:基于命令持久性预测的自适应语义推理以实现高效导航视觉-语言-动作(VLA)

ASPEN: Adaptive Semantic Reasoning via Command Persistence Prediction for Efficient Navigation VLAs

University of Houston(休斯顿大学)

Vieri N. Akustia, Bin Hu

AI总结:

提出ASPEN调度器,通过预测命令持久性实现导航VLA的语义推理按需触发,减少VLA调用与推理时间,同时维持导航性能与基线动作序列一致性。

摘要:

导航视觉-语言-动作(VLA)策略在后续命令未发生变化时仍会重复调用开销高昂的语义推理。命令复用可减少计算量,但存在遗漏指令相关转向或停止的风险。我们提出ASPEN,一种基于命令持久性预测的事件触发调度器:估计冻结的基础策略是否会重新发出其先前的命令。轻量级多模态预测器结合时间观测、语言指令、执行历史和缓存的VLA置信度。在符合条件的命令边界处,若预测的持久性足够则允许命令复用;否则,ASPEN会触发VLA进行新的语义推理。在跨数据集RxR-CE val-unseen上使用冻结的NaVILA进行评估,ASPEN将VLA调用次数减少9.5%,估计推理时间减少8.1%(含调度开销),同时保持相当的导航性能,且在92.1%的回合中保留与基线相同的动作序列。

英文摘要:

Navigation vision-language-action (VLA) policies repeatedly invoke costly semantic reasoning even when the next command remains unchanged. Command reuse can reduce computation but risks missing instruction-dependent turns or stops. We introduce ASPEN, an event-triggered scheduler based on command persistence prediction: estimating whether the frozen base policy would reissue its previous command. A lightweight multimodal predictor combines temporal observations, language instructions, execution history, and cached VLA confidence. At eligible command boundaries, sufficient predicted persistence permits command reuse. Otherwise, ASPEN triggers fresh semantic reasoning from the VLA. Evaluated with frozen NaVILA on cross-dataset RxR-CE val-unseen, ASPEN reduces VLA calls by 9.5% and estimated inference time by 8.1%, including scheduling overhead. It maintains comparable navigation performance and preserves identical baseline action sequences on 92.1% of episodes.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑