允中 发自 凹非寺 量子位 | 公众号 QbitAI 12月17日，在香港举办的全球图形学领域备受瞩目的顶级学术盛会SIGGRAPH Asia 2025上， 摩尔线程 在3D Gaussian Splatting Reconstruction Challenge （3DGS重建挑战赛） 中凭借自研技术 LiteGS 出色的算法实力和软硬件协同优化能力，斩获大赛 银奖 。 这再次证明了，摩尔线程在新一代图形渲染技术上的深度积累与学术界的高度认可。 3DGS：下一代图形渲染的范式革命，开启AI加速的高效渲染时代 3D Gaussian Splatting （3DGS，三维高斯溅射） 是在2023年被提出的一项革命性3D场景表示与渲染技术，以可参数化的3D高斯分布为 核心，实现了画质、效率与资源占用之间的卓越平衡。 与传统NeRF相比，3DGS在保持逼真渲染质量的前提下，将渲染效率提升数百至上千倍，并在 光线追踪、VR/AR实时渲染、多模态融合 等 方向展现出极强的适应性与扩展性。 △ 上图仅作示意 作为近年来快速发展的神经渲染技术，3DGS不仅在三维重建与实时渲染等方向展现出卓越优势，也在更广泛的AI场景中具备潜 ...

Core Insights - The article discusses the advancements in Embodied AI, particularly through the introduction of RoboTidy, which utilizes 3D Gaussian Splatting (3DGS) technology to create realistic interactive 3D environments for training robots [4][8][20]. Group 1: Importance of 3DGS - Embodied AI research has been hindered by the "simulation paradox," where traditional 3D modeling methods result in low-fidelity environments that do not accurately represent real-world complexities [7]. - RoboTidy's breakthrough lies in its use of 3DGS, which allows for high-speed rendering (over 100 FPS) of photorealistic scenes, enhancing the training environment for robots [9][11]. - The research team scanned 500 real household scenes, enabling robots to experience realistic lighting and textures, which significantly improves the robustness of visual encoders [11][12]. Group 2: Redefining Organization Tasks - Organizing a room is a complex long-horizon planning challenge for robots, requiring semantic understanding and common-sense reasoning [14]. - RoboTidy provides a large dataset of over 8000 expert demonstration trajectories, capturing the implicit logic of how humans organize spaces [14][15]. - The framework includes a "Semantic Planner" and "Low-level Policy," allowing robots to learn organization tasks in a human-like manner [15]. Group 3: Sim-to-Real Validation - The collaboration with Yuanli Infinite focuses on bridging the Sim-to-Real gap, addressing a significant industry challenge [17]. - Experiments show that models trained in RoboTidy's high-fidelity environment achieve a task success rate improvement of 29.4% in real-world robot tests compared to traditional methods [17][18]. - This demonstrates that high-quality simulation data can be effectively translated into real-world productivity [18]. Group 4: Standardization and Open Source - Prior to RoboTidy, there was a lack of standardized evaluation metrics for household organization tasks, making it difficult to compare results across different research labs [20]. - RoboTidy establishes a standardized evaluation system and leaderboard, inviting global developers to contribute to the evolution of household service robots [20][22]. - The initiative aims to create a more realistic and rigorous starting point for advancing the field of Embodied AI [22][27].

Core Insights - The article discusses the advancements in Embodied AI, particularly focusing on the RoboTidy project, which aims to enhance the capabilities of robots in household tasks through realistic training environments [3][4]. Group 1: Introduction to RoboTidy - RoboTidy is the first benchmark based on 3D Gaussian Splatting (3DGS) technology, creating 500 photo-realistic interactive 3D environments and providing over 8000 expert demonstration trajectories [4]. - The project demonstrates significant potential in real-world applications, with a nearly 30% increase in task success rates for real robots after training in the RoboTidy environment [4][16]. Group 2: Importance of 3DGS - Traditional simulation environments often suffer from low fidelity, which hampers the performance of robots in real-world scenarios [7]. - 3DGS offers high rendering speeds (over 100 FPS) and realistic scene reconstruction, addressing the limitations of previous methods [8][10]. Group 3: Redefining Organization Tasks - Organizing a room is a complex long-horizon planning challenge for robots, requiring semantic understanding and common-sense reasoning [13]. - RoboTidy provides a large and high-quality dataset that captures the implicit logic of human organization, enabling robots to learn effective planning strategies [14]. Group 4: Sim-to-Real Validation - The collaboration with Yuanli Infinite focuses on bridging the Sim-to-Real gap, a critical industry challenge [16]. - Experiments show that models trained in the RoboTidy environment outperform traditional methods, especially in handling unseen objects and complex backgrounds, with a task success rate improvement of 29.4% [16][17]. Group 5: Standardization and Open Source - RoboTidy establishes a standardized evaluation system and leaderboard, addressing the lack of uniform assessment criteria in household organization tasks [19]. - The project invites global developers to contribute to advancing household service robots on a more realistic and rigorous platform [21]. Group 6: Conclusion - The emergence of RoboTidy signifies a paradigm shift in Embodied AI research, emphasizing the need for stronger algorithms and more realistic environments [23]. - The collaboration between industry and academia, exemplified by Yuanli Infinite and top academic institutions, is seen as a catalyst for the evolution of general-purpose humanoid robots [23][24].