T-RO Survey Disentangles VLA, Diffusion Policy and LLM Planner Axes in Robot Manipulation
jiqizhixin · x · 2026-09-13
A survey from Xi'an Jiaotong University, HKUST(GZ), and Peking University, accepted to IEEE Transactions on Robotics, reorganizes foundation-model-era robot manipulation from planning and learning perspectives. It argues that VLA, Diffusion Policy, Imitation Learning, and LLM Planner are compared side by side but live on different axes: VLA is about how multimodal inputs enter the action model, Diffusion Policy about how actions are generated, Imitation Learning about learning from demonstrations, and LLM Planner about high-level task planning. The paper disentangles four routinely conflated dimensions: system hierarchy, model architecture, action generation method, and learning strategy.
More from Embodied
- Walden Robotics CTO: general-purpose robot manipulation is having its 'GPT moment' — adnothing · 2026-10-30
- Gym-Anything Team Stopped at 25 Simulations Due to Resource Limits, Seeks Collaborators — wellecks · 2026-10-30
- Gym-Anything Adds Audit Agents That Try to Break Simulations With Legal Robot Actions — wellecks · 2026-10-30
- NVIDIA researcher: train robot policies with privileged sim info, improving Pi 0.5 — wellecks · 2026-10-30
- Leveraging Simulator Privileged Information: Three Methods to Improve Robot Models — wellecks · 2026-10-30
- CMU Auto-Builds 25 Robot Simulations with Agents; Frontier Model Solves 0 of 25 Test Tasks — wellecks · 2026-10-30