Quanting Xie

Origami Robotics blog post discussing how high-ratio gearboxes limit robot dexterity, and describing Origami's approach (low-ratio transmissions, axial flux motors) to improve force transparency and sim-to-real transfer.

Proposes Affordance RAG, a hierarchical multimodal retrieval framework constructing affordance-aware embodied memory to improve mobile manipulation retrieval and execution, reporting high retrieval and task success rates in indoor experiments.

Describes a micro-scale axial-flux permanent-magnet motor using PCB-based high-density interconnect windings to achieve high torque density in a thin package suitable for quasi-direct-drive actuation.

Introduces Embodied-RAG, integrating topological maps and a hierarchical semantic forest to provide non-parametric memory for embodied agents, improving retrieval and generation across navigation and explanation queries in large environments.

Presents methods for object-goal navigation in outdoor and unmapped environments using LLM-based reasoning about unseen areas and introduces the OUTDOOR task and evaluation formalism.

A survey and meta-analysis on applying foundation models to robotics, summarizing the state of the field and directions for general-purpose robotic systems.