Ant Group 旗下 Robbyant 开源触觉基础模型 LingBot-Touch
Ant Group's Robbyant Open-Sources LingBot-Touch Tactile Model for Robot Hands
做真机操作的开发者可以关注:Robbyant 把磁性、视触觉、压阻三种传感器统一进一个模型,官方数据在 26 个任务上给四个动作模型带来 14 到 26 点提升。权重开源,直接拿去给自己的灵巧手微调就行,不用从零造触觉数据管线。
蚂蚁集团旗下具身智能团队 Robbyant 开源了触觉基础模型 LingBot-Touch,统一处理磁性、视触觉和压阻三类传感器信号。官方报告称,在 26 个真机任务上,接入该模型后四个不同动作模型的性能提升 14 到 26 个百分点。
技术上,LingBot-Touch 的做法是把异构触觉数据映射到统一的表征空间,使一套触觉模型可以服务不同硬件方案的机械手,而不必为每种传感器单独训练。开源后,团队可直接在 GitHub/HuggingFace 上获取权重,结合自己的灵巧手硬件做下游微调。
触觉是当前灵巧手操作的主要瓶颈之一,此前该方向多为学术工作或专有闭源方案,主流开源生态集中在视觉与语言模态。Robbyant 此番以开源方式切入,与 Google DeepMind 等在触觉表征上的研究形成对照,对做真机操作的机器人团队而言提供了一个可微调的触觉底座选项。
Ant Group's Robbyant Open-Sources LingBot-Touch Tactile Model for Robot Hands
Robbyant's open-source LingBot-Touch unifies magnetic, visuotactile and piezoresistive sensing; the company reports 14 to 26 point gains for four action models on 26 real-robot tasks.
This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include LingBot-Touch, Robbyant, 触觉感知, 开源模型.
Primary source verification: published directly by pandaily with first-party authority.