Limx Dynamics

LimX

LimX Dynamics, a cutting-edge robotics company based in Shenzhen, China, has made significant strides in the field of humanoid robotics with their flagship model, the CL-1. Founded in 2022, LimX Dynamics has quickly established itself as a leader in developing advanced motion control algorithms and leveraging artificial intelligence to create versatile humanoid robots.

Description

LimX Dynamics' humanoid robot, initially prototyped as CL-1 and evolved into the LimX Oli model, represents a significant advancement in embodied intelligent robotics from the Shenzhen-based company founded in 2022. Standing at 165 cm tall and weighing 55 kg, LimX Oli features 31 degrees of freedom (DoF): 2 in the head, 7 per arm (14 total), 3 in the waist, and 6 per leg (12 total), enabling fluid, human-like motions far surpassing the original prototype's 20 DoF and 157 cm height.7034 The architecture emphasizes modular design with hollow, high-torque-density actuators (up to 150 Nm joint torque), allowing ultra-stable full-body balance, complex movements like lying-to-standing transitions, waist rotations beyond human limits, and dynamic navigation over debris, sand, rocks, and stairs using real-time terrain perception.3233 Perception is powered by multi-sensor fusion: proprietary 6-axis IMUs, dual Intel RealSense D435i depth cameras (head and chest-mounted), with extensive interfaces for additional LiDAR, cameras, or sensors via USB 3.0/3.2, 1G Ethernet, and power ports (24V 5A/12V 5A). This enables active perception for whole-body loco-manipulation tasks like autonomous tennis ball pickup/toss, heavy object loading in simulated warehouses, and museum performances.7069 AI integration focuses on reinforcement learning (RL) via NVIDIA Isaac Sim, MuJoCo, and Gazebo simulators, with zero-gap URDF for Sim2Real transfer. Proprietary frameworks like LimX DreamActor and VGM (VideoGenMotion) blend real-world, simulated, and video data for faster training. The modular SDK supports Python, high/low-level control, OTA motion library updates, teleoperation data collection, and onboard LLM deployment for natural interaction and reasoning. Whole-body control algorithms ensure stability during disturbances, running (up to 5 km/h), and manipulation (3 kg per arm).3344 Real-world deployment history includes prototypes since 2023: CL-1 demos of stair climbing, warehouse tasks (2024), evolving to Oli's 2025 public debut at World Robot Conference. Primarily R&D platform (EDU/Lite/Super editions, preorders from $22k), with demos at CCTV gala, Sanxingdui Museum, and construction sites showcasing stability. No widespread commercial deployments yet, but targeted for manufacturing, logistics, households via general-purpose form factor. Funding ($27.5M+) supports scaling toward AGI via embodied AI.3269 (Word count: 612)

Key Features

Full-Body Dynamic Motion

31 DoF with hollow high-torque actuators enable human-like agility, stair climbing, running, and recovery from falls or lying positions.

Multi-Sensor Perception

Proprietary IMUs and dual Intel RealSense D435i cameras for real-time terrain adaptation and loco-manipulation; expandable with external sensors.

Modular Hardware & SDK

Quick-swap end-effectors, arms; Python SDK, OTA updates, Sim2Real support for rapid AI development and deployment.

Reinforcement Learning Optimized

DreamActor framework integrates multi-source data; compatible with NVIDIA Isaac Sim for efficient embodied AI training.

Warehouse & Rough Terrain Ready

Continuous heavy lifting (3kg/arm), debris navigation, demonstrated in simulated industrial environments.

Specifications

Availability	Prototype
Nationality	China
Website	https://www.limxdynamics.com/en
Degrees Of Freedom, Overall	20
Height [Cm]	157
Manipulation Performance	2
Navigation Performance	2
Max Speed (Km/H)	4
Strength [Kg]	18
Weight [Kg]	50
Runtime Pr Charge (Hours)	3
H.G Skill Score	4
Verified	Not verified
Walking Speed [Km/H]	2.3
Manufacturer	Limx Dynamics
Height Cm	165
Weight Kg	55
Dof Overall	31
Dof Head	2
Dof Arm Single	7
Dof Waist	3
Dof Leg Single	6
Max Speed Kmh	5
Max Joint Torque Nm	150
Payload Arm Kg	3
Sensors	Proprietary 6-axis IMU (multiple), Intel RealSense D435i depth cameras (head x1, chest x1)
Interfaces	USB 3.0/3.2, Ethernet 1000 Mbps RJ45, Power 24V 5A / 12V 5A
Actuators	Hollow high-torque-density motors
Materials	Modular lightweight structure (specifics N/A)
Processors	N/A (supports onboard LLM deployment)
Battery	N/A (estimated 2-3 hours runtime)
Os	Linux-based with Python SDK
End Effectors	Modular (2-finger grippers, 5-finger dexterous hands)

Curated Videos

Frequently Asked Questions

What are the main specifications of LimX Oli?

LimX Oli measures 165 cm in height, weighs 55 kg, and has 31 DoF (excluding end-effectors). Max speed is 5 km/h, joint torque up to 150 Nm, payload 3 kg per arm. Equipped with dual Intel RealSense D435i cameras and proprietary 6-axis IMUs for perception.

What AI technologies does it use?

Utilizes reinforcement learning with NVIDIA Isaac Sim, LimX DreamActor for multi-source data training (real, sim, video), and VGM framework. Supports onboard LLMs, whole-body control, and Sim2Real transfer for advanced perception and motion planning.

Is it suitable for real-world deployment?

Primarily an R&D platform (EDU edition), with demos in warehouses, museums, and rough terrain. Modular design and OTA updates facilitate transition to industrial tasks like logistics and manufacturing, though commercial scale-up is ongoing.

What development tools are available?

Open modular SDK with Python support, high/low-level interfaces, predefined motion libraries (OTA-updatable), URDF models for simulation (Isaac Sim, MuJoCo), teleoperation for data collection, and GitHub resources for continuous updates.

How does it handle uneven terrain?

Real-time terrain perception via depth cameras and IMUs enables dynamic stair climbing, debris navigation (sand, rocks), and balance recovery. Closed-loop system integrates perception, gait planning, and control for proactive adjustments.

What is the battery life or power details?

Specific battery capacity not publicly detailed; runtime estimated 2-3 hours based on prototypes. Power interfaces include 24V 5A and 12V 5A for external charging and peripherals.