YC F26 批次初创公司 Axonnic 推出了一款前臂袖套,旨在通过捕捉人类精细动作背后的肌肉努力来生成高质量训练数据,辅助灵巧机器人学习 1。该产品利用 64 通道表面肌电(sEMG)技术读取肌肉活动以解码手指运动并估算指尖力度,同时同步手部追踪训练模型预测关节角度,其设计核心在于保持使用者手部裸露 1。这一技术方案源自创始人之一 Max 在牛津大学进行的肌电(EMG)研究,团队成员 Arsh 和 Max 均具备牛津大学的学术背景及工业界经验 1。
在性能测试方面,Axonnic 展示了其在工厂环境下的实测成果:在一次未进行重新校准的训练中,系统录制了 215 秒的数据 1。将录制结果与 Apple Vision Pro 设备数据进行对比时,该系统对 20 个关节角度的预测误差仅为 6.9° 1。除了技术验证,公司在商业化进展上也取得突破,已与一家欧洲制造商签署意向书(LOI),用于缝纫等手工任务的机器人实施支持 1。此外,团队已收集建筑工地的演示数据,并正在开发基于有线 PCB 设计的自有电子设备,以消除无线干扰和时序不一致 1。
Axonnic, a startup from the Y Combinator F26 batch, has launched forearm sleeves designed to track finger movements and estimate fingertip force by reading muscle activity 1. The device utilizes 64-channel surface electromyography (sEMG) technology while keeping the hands bare, allowing it to decode fine motor skills and generate high-quality training data to assist in teaching dexterous robots 1. By capturing the underlying muscular effort behind human actions, the system aims to improve robotic learning processes through synchronized hand tracking and joint angle prediction models 1.
During testing, the team demonstrated significant accuracy without needing recalibration; in one session recording 215 seconds of data, the system showed an error margin of only 6.9° across 20 joint angles when compared against Apple Vision Pro standards 1. Commercially, Axonnic has signed a Letter of Intent with a European manufacturer to support robot deployment in manual tasks such as sewing and has already collected demonstration data from construction sites 1. To address technical challenges like wireless interference and timing inconsistencies, the company is currently developing its own hardware based on wired PCB designs 1. The founders, Arsh and Max, bring academic and industrial experience from Oxford University, where the project originated from Max’s research on EMG technology 1.
评论
还没有评论,欢迎留下第一条。