technique

diffusion-based policy

techniqueactivediffusion-based-policy-52477a9f·1 events·first seen 26d ago

Aliases: diffusion-based policy

Co-occurring entities

More like this (12)

Diffusion Policy Ambient Diffusion Policy Denoising Diffusion Policy Optimization Diffusion Models policy gradient diffusion-based generative models diffusion-based inpainting latent diffusion model Policy Gradient Methods Masked Diffusion Models latent diffusion Representation-Conditioned Diffusion Models

Recent events (1)

5arXiv · cs.AI·26d ago·source ↗

HITL-D: Human-In-The-Loop Diffusion for Shared Control in Robotic Manipulation

HITL-D is a shared control framework that combines diffusion-based policies with human teleoperation for robotic manipulation tasks. The system autonomously updates end-effector orientation conditioned on scene point clouds and Cartesian position, reducing the number of joystick axes operators must manage. A 12-participant user study found 40% faster task completion, 37% lower perceived workload, and improved subjective ratings versus traditional teleoperation. The work addresses a relatively unexplored intersection of diffusion policy methods and human-in-the-loop control.

Agent and Tool Ecosystem Alignment and RLHF HITL-D diffusion-based policy shared control +1 more