Event
PAL hierarchical memory framework paper released and evaluated
1 reports · 1 sources
Qinghang Liu et al. released a paper on arXiv proposing PAL (Plan first, Act Later), a hierarchical memory framework for vision-language robotic manipulation, and completed evaluations on RoboMME and real-world manipulation tasks; code and videos are publicly available on the authors' website. In this framework, a high-level vision-language model (VLM) first converts streaming visual observations into a subtask_list, and during execution selects the currently relevant subtask from it to hand to a low-level vision-language-action (VLA) policy for execution.
Report timeline
- PAL: A Hierarchical Memory Framework with Plan-First, Act-Later Streaming Reasoning for Robot Manipulation
The paper proposes PAL (Plan first, Act Later), a hierarchical memory framework for vision-language robot manipulation. A high-level vision-language model (VLM) converts streaming visual observations into a subtask_list, and at execution time selects the currently relevant subtask from it to pass to a low-level vision-language-action (VLA) policy.
arXiv cs.RO ·