China Unveils X2-N Humanoid Robot with Wheels-for-Feet Mobility

Chinese robotics company AgiBot has introduced the X2-N, a humanoid robot capable of switching between walking and wheeled movement to navigate complex terrain.

The X2-N uses a hybrid locomotion system that allows it to climb stairs in bipedal mode and traverse ramps or narrow bridges in wheeled mode. It can carry loads up to 5.4 kg while walking.

Unlike traditional robots that rely on cameras or external sensors, the X2-N uses proprioception—internal data such as joint torque, pressure feedback, and gyroscopic input—to sense and adapt to its environment in real time.

This sensor-less navigation mimics human intuition, enabling the robot to adjust posture and movement without visual input. The approach reduces system complexity, eliminates latency, and enhances robustness.

A demonstration video shows the X2-N rolling across flat ground, switching to walking mode to climb stairs, then returning to wheels to ascend a slope, highlighting its terrain adaptability.

The X2-N is based on AgiBot’s Lingxi X2, a 1.3-metre-tall humanoid robot weighing 33.8 kg. Introduced in March 2025, the Lingxi X2 blends AI and robotics, capable of walking, running, dancing, turning, scooter riding, and cycling.

Lingxi X2 features a multimodal interaction system that responds within milliseconds by analysing facial expressions and vocal tones to detect emotional states and deliver context-aware responses.

Its design includes simulated breathing, subtle body language, and passive observation, enhancing emotional expressiveness. It can also perform basic object manipulation tasks without prior training.

These capabilities are powered by AgiBot’s Genie Operator-1 (GO-1), an AI model that supports zero-sample generalisation and multi-robot coordination. GO-1 uses latent actions and a transformer-based Latent Planner to improve decision-making and motion execution.

In benchmark tests across five tasks, GO-1 increased robot success rates from 46% to 78%, with notable improvements in complex tasks like pouring and refilling. The Latent Planner alone improved planning efficiency by 12%.

Potential applications for the X2-N include logistics, healthcare, and public services. Future versions may integrate visual input with proprioception to enhance situational awareness.

---

• AgiBot’s X2-N robot switches between walking and wheeled movement

• Uses internal sensors for real-time terrain adaptation without cameras

• Based on Lingxi X2, a 1.3-metre humanoid with emotional interaction features

Source: INTERESTING ENGINEERING