Chinese firm Aimoga Robotics has demonstrated its Mornine humanoid robot autonomously opening a car door inside a working Chery dealership using only onboard sensors, full-body motion control and end-to-end reinforcement learning.

Unlike traditional scripted robots or teleoperated systems, Mornine completes the task without human intervention. The robot identifies the door handle, adjusts its posture and pulls open the door using coordinated force across its arms, waist and legs. The deployment marks one of the first instances of a service robot executing such a high-friction, physical interaction in a live commercial setting.

Mornine’s sensor stack includes 3D lidar, depth and wide-angle cameras, and a visual-language model (VLM), allowing it to perceive door position and opening status in real time. Rather than being programmed to recognise handles, the robot learned this through reinforcement learning: over millions of simulation cycles, the model independently learned to focus on the correct region and execute the task.

“We never explicitly told the robot what a door handle is,” said the engineering team. “It learned to focus on that region by itself.”

After simulation training, the learned model was deployed via Sim2Real methods. Mornine continues collecting live sensor data, which feeds back into the cloud training loop for ongoing improvement, enabling a continuous learning cycle in real deployments.

Mornine is operating in several Chery (www.cheryinternational.com) 4S dealerships in China. Beyond opening car doors, it assists in customer greeting, vehicle introduction and item delivery. Its ability to handle physical tasks such as door opening supports Aimoga’s vision of humanoid robots working alongside humans in real-world retail environments.

“Opening a car door may seem simple,” the company noted, “but in robotics, it marks a shift: from simulation to service, from command to capability.”