We walk without worrying about what’s behind us. Why should humanoids?
Overview
Source roundup from published reports. Claims below are attributed to their publishers, not independently verified. ETH Robotic Systems Lab — X: Watch Your Step: Anticipated Perception for Humanoid Locomotion. The ETH Robotic Systems Lab introduces a new perception approach for humanoid robots navigating rough terrain, focusing on predicting and perceiving only the terrain patches relevant for the next step, aiming to enhance … ETH Robotic Systems Lab — X: We walk without worrying about what’s behind us. Why should humanoids?. The ETH Robotic Systems Lab introduces 'Anticipated Perception for Humanoid Locomotion,' a new method that prioritizes terrain perception for the feet's upcoming landing points, aiming to enhance stability and navigation…
Generated from attributed reports · Updated 43 minutes ago
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · Reported success rate: 59 percent · Basis not reported · Differing source assertions
“在239条长距离闭环轨迹的测试中，DA-Nav的导航成功率达到59.00%，偏差出现后的纠偏成功率高达98.15%，相比之下，其他对比方案的纠偏成功率普遍只有个位数。”
Exact source · revision 1robospeak
Reported quantity · Reported success rate: 98.15 percent · Basis not reported · Differing source assertions
“在239条长距离闭环轨迹的测试中，DA-Nav的导航成功率达到59.00%，偏差出现后的纠偏成功率高达98.15%，相比之下，其他对比方案的纠偏成功率普遍只有个位数。”
Exact source · revision 1robospeak
Reported quantity · navigation distance: 1200 other · Basis not reported
“团队把训练好的模型直接部署在厂商四足机器人和人形机器人上，没有在真实数据上做任何微调，就在不同城市环境中完成了超过1200米的稳定导航。”
Exact source · revision 1robospeak
Reported quantity · number of trajectories: 239 other · trajectory count
“在239条长距离闭环轨迹的测试中，DA-Nav的导航成功率达到59.00%，偏差出现后的纠偏成功率高达98.15%，相比之下，其他对比方案的纠偏成功率普遍只有个位数。”
Exact source · revision 1robospeak
Reported quantity · number of sequences: 286000 other · Basis not reported
“整个数据集包含约28.6万序列样本，其中约12.8万帧是恢复行为数据，与15.8万帧的正常导航数据混合训练后，模型学会了面对偏离时如何拉回正轨。”
Exact source · revision 1robospeak
Reported quantity · number of frames: 128000 other · Basis not reported · Differing source assertions
“整个数据集包含约28.6万序列样本，其中约12.8万帧是恢复行为数据，与15.8万帧的正常导航数据混合训练后，模型学会了面对偏离时如何拉回正轨。”
Exact source · revision 1robospeak
Reported quantity · number of frames: 158000 other · Basis not reported · Differing source assertions
“整个数据集包含约28.6万序列样本，其中约12.8万帧是恢复行为数据，与15.8万帧的正常导航数据混合训练后，模型学会了面对偏离时如何拉回正轨。”
Exact source · revision 1robospeak
Evidence dependency changes (3)
- 2026-10-30T17:22:21.927Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:19:09.323Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:19:01.360Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
Developments
- 2026-10-29 09:56 UTC · 1 reportsWatch Your Step: Anticipated Perception for Humanoid LocomotionETH Robotic Systems Lab — X：Watch Your Step: Anticipated Perception for Humanoid Locomotion
- 2026-10-29 09:44 UTC · 1 reportsWe walk without worrying about what’s behind us. Why should humanoids?ETH Robotic Systems Lab — X：We walk without worrying about what’s behind us. Why should humanoids?
Report timeline
Follow attributed reports and material updates.
- ETH Robotic Systems Lab — XSignalWatch Your Step: Anticipated Perception for Humanoid Locomotion
The ETH Robotic Systems Lab introduces a new perception approach for humanoid robots navigating rough terrain, focusing on predicting and perceiving only the terrain patches relevant for the next step, aiming to enhance locomotion efficiency and adaptability.
- ETH Robotic Systems Lab — XSignalWe walk without worrying about what’s behind us. Why should humanoids?
The ETH Robotic Systems Lab introduces 'Anticipated Perception for Humanoid Locomotion,' a new method that prioritizes terrain perception for the feet's upcoming landing points, aiming to enhance stability and navigation in humanoids.
- RoboSpeak — WeChatSignalThe 'Double Helix' of Embodied Intelligence, Happening in Space?
The StarSource Intelligence team, in collaboration with universities and research institutions, has launched two studies, DA-Nav and AdaOcc, addressing the compatibility issues between robot navigation and three-dimensional perception. DA-Nav achieves city-level navigation through discrete image point selection and chain-of-thought reasoning, while AdaOcc enhances the flexibility and robustness of three-dimensional perception through adaptive geometric guidance and sparse point representation. These two works demonstrate key technical pathways for embodied intelligence in real-world environments.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).