Zerith H1 Shows Fully Autonomous Cleaning in Busy Mall
Zerith Robotics has officially taken its H1 humanoid robot out of the lab and into a real commercial environment. The company recently showcased a live deployment at the Hefei Vientiane City shopping mall, where the H1 operated autonomously across multiple cleaning workflows, from mopping floors to wiping sinks and maintaining restrooms.
It is one of the clearest real-world demonstrations yet of humanoids performing practical, repetitive work in high-traffic public spaces.
A Humanoid Built for Real Work, Not Demos
Unlike controlled lab demos or staged tests, the H1 ran tasks end-to-end on its own inside an active mall. Zerith says the robot handled:
Mopping and surface cleaning
Restroom sink and counter wiping
General sanitation maintenance
Autonomous navigation through public areas
The deployment is part of Zerith’s push to show that humanoids can reliably perform service-industry tasks that require dexterity, mobility, and environmental awareness.
Powered by Zerith-V0: A Cognitive + Behavioral Architecture
At the heart of the H1’s capabilities is Zerith-V0, the company’s new foundation model designed specifically for embodied robots.
Zerith describes the system as a “Cognitive-Behavioral” dual-system architecture, separating object understanding from action execution:
Cognitive system: Interprets objects, surfaces, spatial layout, and task requirements
Behavioral system: Handles motion planning, manipulation, and control
The separation is designed to help the robot focus on relevant tasks while ignoring environmental distractions, a crucial requirement in noisy, dynamic public venues like malls.
Training With Real2Sim2Real and Self-Correcting RL
Zerith’s training pipeline is built around a Real2Sim2Real loop, using real-world demonstrations to refine simulated environments and push improvements back into the physical robot.
On top of that, the H1 uses Self-Correcting Reinforcement Learning, an approach where mistakes made in real-world deployment are fed back into the training cycle so the robot can autonomously improve over time.
This combination aims to boost reliability and reduce failure rates in messy, unpredictable settings, exactly the environments humanoids must conquer to reach commercial scale.
Why This Deployment Matters
Humanoid companies worldwide are racing to prove that robots can provide meaningful labor in customer-facing and service-heavy industries. Zerith’s mall deployment demonstrates:
Real-world autonomy with minimal human oversight
Fine-motor cleaning tasks traditionally done by humans
Operational stability in crowded public areas
A training pipeline designed for continuous improvement
As humanoids expand beyond factories and into public infrastructure, malls, airports, hotels, hospitals, deployments like this signal how quickly the sector is maturing.
Zerith’s H1 may be one of the earliest humanoids to handle restroom cleaning autonomously in a commercial space, a milestone that shows how close real, revenue-generating humanoid labor is becoming.
