This humanoid robot can lift heavy objects, autonomously replan tasks

There’s a growing need for robots that can perform difficult tasks accurately. In this arena, LimX Dynamics has showcased its humanoid CL-1 robot that can successfully lift heavy objects.

The robot can help around the house, and perform multiple tasks, according to the company.

In a recent video shared by LimX Dynamics, the robot is seen loading heavy objects among shelves in a simulated warehouse.

In the three-minute clip, CL-1 is seen adopting multiple postures, including loaded deep squat, loaded stand from a squat, loaded lift, and loaded walking.

Humanoid CL-1 robot performed multiple tasks

The robot was able to perform these tasks after the company enhanced whole-body motion control and real-time perceptive loco-manipulation with hardware upgrades.

The humanoid robot can now filter out unpredictable external disturbances, re-locate the target object, autonomously replan the task, and adjust its position to place the object at the designated location.

Now, the robot maintains stable performance during multiple mobile operations, addressing the challenges of long-duration work tasks.

CL-1 robot achieved dynamic stair climbing

The company earlier claimed that CL-1 is one of the few humanoid robots around the world that achieves dynamic stair climbing based on real-time terrain perception. As an ideal testing platform for Embodied AI, CL-1 has taken a significant step forward to become the next widely used intelligent terminal besides cars and smartphones in the future.

Through closing the loop between real-time perception, gait planning, and locomotion control, and hardware and data stream, CL-1 dynamically completed tasks such as climbing stairs, walking down slopes, and, locomoting indoors and outdoors in different environments.

LimX Dynamics claimed that it has developed a real-time, distributed operation system, including three key parts. Integrated development tool chain greatly reduces coding complexity and improves the efficiency of algorithm development.

Real-time data transmission makes accurate operation of the robot possible. Decentralized operation of the nodes secures a balanced computing power, achieving a more flexible, efficient, and stable system. It also supports a wide range of programming languages to maximize the diverse technical requirements.

LimX Dynamics claimed to have mastered the leading technologies including model predictive control, reinforcement learning, high frequency torque control.

The company claims that it has comprehensive grasp of the robot’s intrinsic physical movement capabilities. Based on the terrain information centered on the robot’s body and its relative position, combined with the real-time state of the robot including its location, position, orientation, joint angles, etc.

“We compute the optimal landing point, the magnitude of interaction forces, and the velocity of movement, achieving the best interaction between the robot and the terrain,” said LimX Dynamics.