# 面向人形机器人非周期步态的多模态跌倒预测新架构与基准

> 原文链接：https://robotics.hellobike.com/papers/2026-11-20-2609-07930-f2bddc98
> 频道：papers | 更新日期：2026-11-20 | 标签：auto, arxiv
> 核心摘要：针对人形机器人非周期运动跌倒预测难题，提出融合视觉与本体感知的多模态深度学习架构及两套新基准。实测数据性能提升超12.73%，仿真提升超10.40%，有效解决传统时序模型在非周期轨迹下的失效问题。

# 面向人形机器人非周期步态的多模态跌倒预测新架构与基准

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[A Multimodal Label Forecasting Method for Aperiodic Visuo-Motor Time Series](https://arxiv.org/abs/2609.07930)（arXiv 2609.07930）

## 核心内容

针对人形机器人运动轨迹多样性导致周期性假设失效、现有TSF模型难以准确预测跌倒的问题，研究团队发布了包含仿真与真机数据的两套非周期视动时间序列基准。提出的新型多模态标签预测架构在真实硬件数据集上取得12.73%以上的显著性能提升，仿真环境提升超10.40%，验证了其在非平稳序列下的有效性。

该架构联合处理自我中心视角视觉与本体感知信号，利用内源与外生变量建模非周期动态。训练流程严格执行独立同分布采样以消除时序偏差，克服了Transformer与MLP类模型对周期性的依赖。代码与数据集将开源，为非周期具身智能任务提供了标准化的评测体系与算法范式。

## 关键看点

- **非周期基准**：发布仿真与真机双数据集，证实人形机器人多样步态违反周期性假设，现有深度TSF方法在此场景下性能显著下降。
- **多模态架构**：融合自我中心视觉与本体感知，利用内源及外生变量建模，突破单模态时序预测在非周期运动中的局限。
- **量化性能提升**：在真实硬件数据上性能提升≥12.73%，仿真数据提升≥10.40%，统计学显著优于当前主流时序预测模型。
- **训练策略优化**：采用严格的i.i.d采样机制构建训练样本，消除非周期时间序列中的分布偏移，保障模型泛化能力。

## 工程与产业观察

该工作揭示了具身智能时序预测不能简单套用通用TSF范式，必须针对非周期物理特性重构模型。其开源基准填补了人形机器人跌倒预测的数据空白，强调多模态融合与严格采样策略对提升端侧安全模型鲁棒性的关键作用，为具身控制算法的可靠性验证提供了重要工程参考。

## 英文摘要 (Abstract)

Deep learning models have been increasingly applied to Time Series Forecasting (TSF) in recent years. Transformer-based and MLP-based models have both been used effectively on many real-world TSF regression benchmarks, and there is ongoing debate as to which family of methods is best. While these benchmarks have drawn much attention, it is also worth noting that many current datasets and methods assume approximate periodicity in the time series. In this work, we focus on a new TSF task without periodicity: anticipating falls during humanoid locomotion, on the basis of egocentric vision and proprioception. When the locomotion trajectories are sufficiently diverse, periodicity is violated. We contribute two new benchmark datasets (one from simulation, one from real hardware), showing that periodicity is violated and recent deep TSF methods struggle on these benchmarks. We also propose a novel deep learning architecture that exploits both endogenous and exogenous variables and a training process that rigorously enforces i.i.d sampling of training examples. Our results show statistically significant improvement over prior art in multiple experimental conditions, by 12.73% or more on the real data and 10.40% or more on the simulation data. Code and datasets will be available upon acceptance.

作者：Borui He, Garrett E Katz
:::

## 原始参考来源 (Sources)

- [arXiv 2609.07930：A Multimodal Label Forecasting Method for Aperiodic Visuo-Motor Time Series](https://arxiv.org/abs/2609.07930)