AiMOGA Robotics has achieved a significant breakthrough in embodied artificial intelligence. Its humanoid robot, Mornine, has become the world’s first to autonomously open a car door in a live dealership environment, without human intervention or pre-scripted commands.
The milestone was reached inside a working Chery 4S dealership in China. Using only onboard sensors and reinforcement learning, Mornine identified a car door handle, adjusted its full-body posture, and executed the physical action of pulling the door open. The feat was achieved using coordinated motion across the robot’s arms, waist, and legs — a complex interaction that few service robots have successfully attempted in real-world scenarios.
Unlike traditional robots that follow programmed scripts or are teleoperated, Mornine functions independently. Its sensor stack includes 3D LiDAR, wide-angle and depth cameras, and a visual-language model. These technologies enable real-time perception of the door’s position and status.
The robot learned how to perform the task through a process known as reinforcement learning. Over millions of simulation cycles, it was not explicitly told what a door handle is but learned to identify and focus on it based on task success. “We never explicitly told the robot what a door handle is,” the AiMOGA engineering team said. “It learned to focus on that region by itself.”
Following simulation training, AiMOGA deployed the robot’s learned model into the physical world using Sim2Real techniques. Mornine continues to collect live sensor data during operations, which is sent to the cloud to refine its capabilities through ongoing training — forming a closed feedback loop that allows for continuous improvement in real deployments.
Mornine is already in use in multiple Chery dealerships, where it greets customers, introduces vehicles, delivers items, and now, opens car doors. AiMOGA believes such real-world autonomy signals a shift in robotics — from lab-based simulations to practical service roles.
“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.”
