Cloud Depth Tech Releases Five Major Achievements, Building a Complete Embodied Brain Architecture
云深处科技密集发布五大成果，构筑具身大脑完整版图
Cloud Depth Tech has released five major results, including DR-WM-1.0, DR-VLA-2.0, DR-Agent, DR-APO, and DR-Benchmark, covering the full technical chain of embodied intelligence. These advancements focus on physical constraints, human social rules, strategy optimization, self-evolving navigation, and real-world data collection, marking a significant step toward practical deployment of embodied AI systems.
The release of DR-WM-1.0, DR-VLA-2.0, DR-APO, DR-Agent, and DR-Benchmark represents a comprehensive advancement in embodied intelligence, integrating physical constraints, human-centric learning, and real-world data collection to enhance robot navigation and interaction capabilities.
Source: Chinese robotics — Robot learning research · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
human_egocentric
10,000 hours
View original evidence
团队构建的EgoLynx数据集，整合1万小时人类第一视角网络视频与2000段机器人示教片段，创新利用DISR模块，机器人得以从海量视频中学习行为模式、理解交通规则与导航规则、理解人类的意图、学会社交礼仪，进一步提升具身智能的拟人化程度。
Open source S12Reported success rate
89.55%
View original evidence
在5项MuJoCo-v5连续控制基准测试中，DR-APO在扩散策略上十项指标九项第一、在流策略上十项指标全部第一，取得最佳单策略综合性能；在具身视觉导航后训练中，预训练模型NavDP的仿真成功率从70.33%提升至89.55%；在云深处科技山猫M20 轮足机器人 真实场景测试中，导航成功率从25%大幅提升至65%。
Open source S22Reported success rate
65%
View original evidence
在5项MuJoCo-v5连续控制基准测试中，DR-APO在扩散策略上十项指标九项第一、在流策略上十项指标全部第一，取得最佳单策略综合性能；在具身视觉导航后训练中，预训练模型NavDP的仿真成功率从70.33%提升至89.55%；在云深处科技山猫M20 轮足机器人 真实场景测试中，导航成功率从25%大幅提升至65%。
Open source S22Reported success rate
70.33%
View original evidence
在5项MuJoCo-v5连续控制基准测试中，DR-APO在扩散策略上十项指标九项第一、在流策略上十项指标全部第一，取得最佳单策略综合性能；在具身视觉导航后训练中，预训练模型NavDP的仿真成功率从70.33%提升至89.55%；在云深处科技山猫M20 轮足机器人 真实场景测试中，导航成功率从25%大幅提升至65%。
Open source S22Reported success rate
25%
View original evidence
在5项MuJoCo-v5连续控制基准测试中，DR-APO在扩散策略上十项指标九项第一、在流策略上十项指标全部第一，取得最佳单策略综合性能；在具身视觉导航后训练中，预训练模型NavDP的仿真成功率从70.33%提升至89.55%；在云深处科技山猫M20 轮足机器人 真实场景测试中，导航成功率从25%大幅提升至65%。
Open source S22
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
团队构建的EgoLynx数据集，整合1万小时人类第一视角网络视频与2000段机器人示教片段，创新利用DISR模块，机器人得以从海量视频中学习行为模式、理解交通规则与导航规则、理解人类的意图、学会社交礼仪，进一步提升具身智能的拟人化程度。
Open source S12
在EgoLynx-Robot测试集评估中，导航综合评分QNS提升约5%，最终航向误差降低14.6%。
Open source S16
在5项MuJoCo-v5连续控制基准测试中，DR-APO在扩散策略上十项指标九项第一、在流策略上十项指标全部第一，取得最佳单策略综合性能；在具身视觉导航后训练中，预训练模型NavDP的仿真成功率从70.33%提升至89.55%；在云深处科技山猫M20 轮足机器人 真实场景测试中，导航成功率从25%大幅提升至65%。
Open source S22
Source：Chinese robotics — Robot learning research · news.qq.com