外观
论文arXiv 2610.0759422 天前
BiGym 2.0：面向人形机器人家庭操作的学习策略与智能体开发策略基准测试
BiGym 2.0面向人形家庭操作，统一全身控制器，比较多类策略，并在Unitree G1的20项家务任务上验证。
宇树 G1原文链接：BiGym 2.0: Benchmarking Learned and Agent-Developed Policies for Humanoid Household Manipulation（arXiv 2610.07594 · 发布日期：2026-10-06）
BiGym 2.0: Benchmarking Learned and Agent-Developed Policies for Humanoid Household Manipulation
BiGym 2.0面向人形家庭操作，统一全身控制器，比较多类策略，并在Unitree G1的20项家务任务上验证。
核心要点
- 人形家务操作要求手臂动作同时保持身体平衡、迈步并变换姿态。
- 套件用统一全身控制器，提供每任务60条VR人类演示与多相机记录。
- 在相同视角、本体感知和控制器下，VLA微调取得最高九任务均值。
- 环境、人类演示和评测轨迹已开源；跨工作区堆叠等仍未解决。
验证范围与局限：验证范围为适配Unitree G1的20项家务任务。
英文摘要 (Abstract)
Humanoid household manipulation requires the arms to act while the body balances, steps and changes posture. We present BiGym 2.0, an adaptation of BiGym for the Unitree G1 across 20 household tasks using a unified whole-body controller for demonstration and evaluation. The suite provides 60 native human virtual-reality demonstrations per task with synchronised multi-camera views and full-body execution records. We benchmark vision-language-action fine-tuning, imitation learning, demo-driven reinforcement learning, and cold-start coding agents given the interaction budget of online reinforcement learning. With the same onboard views, proprioception and whole-body controller for every method, vision-language-action fine-tuning has the highest nine-task mean, and agent-developed programs outperform every demo-driven reinforcement learning baseline on this mean and lead on bimanual reaching. Cross-workspace stacking remains open, $π_{0.5}$ stays low on pick-box, and multi-object transport is hard for imitation learning, demo-driven reinforcement learning and coding agents. All environments, human demonstrations, and evaluation traces are open-sourced at https://github.com/swirl-uk/BiGym2.
作者：Zexi Zhang, Zecheng Zhu, Zidong Chen, Zulkhuu Tuya, Stephen James
来源
本页事实对应的公开页面，以原文为准。