N1 Robotics offers Waldo, a teleoperation system that lets operators control humanoid robot hands with motion‑capture gloves and trackers, converting human hand movements intoreal‑time robot manipulation. The platform generates high‑quality manipulation demonstrations for robot learning, and its N1 extension scales data collection for partners while preserving fidelity. Revenue is generated through hardware sales priced at $9,000–$12,000 per unit and associated data‑collection services.
Funding
Funding not disclosed
Founders
Product
Problem
Robotics teams building humanoid manipulators lack a fast, reliable way to capture high‑fidelity hand‑level demonstrations. Existing teleoperation setups are bulky, have low data throughput, and often require multiple operators to reset scenes, which limits the volume and consistency of training data for robot‑learning pipelines.
Solution
N1 offers Waldo, a teleoperation platform that lets a single operator control a humanoid robot’s hands in real time using motion‑capture gloves and body trackers. Raw glove data is processed by WaldoRT, a learned retargeting layer that maps human hand geometry to robot joint commands while preserving grasp topology, finger coordination, and pinch geometry across any hand size or end‑effector. The system delivers up to seven times higher demonstration throughput than consumer‑grade alternatives, producing consistent, high‑quality manipulation episodes that can be directly fed into imitation‑learning or reinforcement‑learning workflows. N1 also provides a managed data‑collection service that scales the same interface for partners, ensuring large‑scale datasets retain the fidelity of the original demonstrations without the quality loss typical of simulation‑based pipelines.
Target Audience
Primary customers are research labs, university robotics groups, and commercial teams developing humanoid manipulation systems that rely on demonstration‑driven learning pipelines.
Features
- Manus‑style motion‑capture gloves combined with inertial trackers delivering sub‑10 ms latency hand‑to‑robot mapping.
- WaldoRT learned retargeting model that automatically adapts human finger angles to robot kinematics, eliminating manual calibration for different hand sizes or end‑effectors.
- Real‑time teleoperation interface supporting seamless reset and rapid re‑attempt of grasps, achieving up to 7× higher episode throughput compared to Meta Quest setups.
- Integrated data pipeline that records synchronized robot joint states, vision streams, and operator intent, exporting datasets in ROS‑compatible formats for downstream learning.
- Scalable data‑collection service that provisions multiple operator stations and aggregates demonstrations into a centralized, version‑controlled repository.
- Secure, end‑to‑end encrypted transmission and storage complying with ISO 27001 standards, with role‑based access controls for research teams.
- API and SDK for easy integration with existing robot control stacks, simulation environments, and cloud‑based training platforms.