NYU's Temporal Flow Policy adds physical time to robot memory, beating RNN and state-space memory designs
newyorkuniversity · hf · 2026-10-15
NYU researchers identify physical time as a missing dimension of robot policy memory: identical observation histories can span different durations and require different actions, yet existing memories treat every update as the same amount of time.
- Proposes Temporal Flow Policy (TFP): memory dynamics that retain past events and evolve the memory state over measured time intervals.
- TFP conditions diffusion and flow-matching policies through this memory state, trained with native imitation objectives; evaluated with Diffusion Policy and π0.5.
- Outperforms recurrent/state-space memories given the same intervals and a learned-gate variant; memory readouts recover elapsed time since events and motion after occlusion.
- Validated on real-robot tasks requiring timing or memory plus LIBERO, suggesting memory should be viewed as a state evolving with physical time, not stored observation history.
More from Embodied
- MimicX refines humanoid tracking with policy-in-the-loop feedback: 25.7% error cut — Shuaijun Liu · 2026-10-15
- Gracia's new world model targets Real2Sim2Real and Physical AI — jonstephens85 · 2026-10-15
- Niantic Spatial built VPS on Pokémon Go imagery; a dev already prototyped it solo — jonstephens85 · 2026-10-15
- RoboParty launches RP1, a fully open-source full-stack humanoid robotics platform — GuanyaShi · 2026-10-15
- Survey: Towards Agentic Robots organizes the field into 17 categories across four domains — Chonghan Qin · 2026-10-15
- HumanoidToolBench benchmarks humanoid tool use from selection to mobile execution on real Unitree G1 — snupilab · 2026-10-15