SPD: VR Simulation Pre-training for Bimanual Dexterity
Loading video
Loading videoSPD from Chelsea Finn's lab and MIT: five operators collect 75 hours of in-simulation VR teleoperation data in one week to pre-train dexterous manipulation policies, then fine-tune on a 56-DoF bimanual setup, beating behavior cloning from scratch.
SPD灵巧操作Dexterous Manipulation双臂BimanualVR遥操作Teleoperation仿真Simulation模仿学习Imitation Learning斯坦福StanfordChelsea FinnMITTwitter
Category: arm
Author: @yokoshaan
Date: 2026-08-23T00:00:00
Duration: 73.873s
Reference: https://arxiv.org/abs/2608.15917
