AI-Powered Robots Show Off Cat-Like Agility at Robotics Conference

Quadruped robot developer Deep Robotics has unveiled its series of AI-enabled robots, which use reinforcement learning for enhanced dexterity, autonomy and self-balancing abilities.

The company’s tech was showcased at this year’s IEEE International Conference on Robotics and Automation (ICRA), taking place last week in Yokohama, Japan.

One of the robots, the Lite3, demonstrated its capabilities by running across a stack of boxes, jumping over gaps and navigating stairs, as well as showing its cat-like ability to readjust its center of gravity when thrown in the air, landing on its feet.

To achieve this, Lite3 was trained using several AI simulations, with the robot’s avatar initially programmed to always be on a slant to strengthen its center of gravity control.

“We successfully applied the Proximal Policy Optimization (PPO) method combined with depth camera sensor inputs and trained an end-to-end policy,” said Zunwang Ma, a robot control engineer at Deep Robotics.

“This approach effectively broadens the limitations of traditional model-based control methods, enabling our quadruped robot to traverse high platforms and gaps without requiring precise perception and robot modeling.”

The company’s larger-scale robot, the X30, was also on display. This model showcased its autonomous capabilities, adapting to a range of terrain, including grass, gravel, and rubble.

Related:Deep Robotics Launches Robot Dog

According to Deep Robotics it is also designed to “withstand external interference” like pushing or pulling. Its ability to maintain balance will be a significant advantage in outdoor operations, the company said. 

Potential applications for the robots include autonomous inspection, emergency detection and rescue.

Moving forward, the company said it’s looking to incorporate advanced AI technologies into its robotic systems, as part of its “AI-Plus Plan” to support and develop innovations in embodied AI. Under this plan, Depp Robotics said it hopes to expand its product capabilities to incorporate perception, planning, decision-making and human-machine interaction.

Ma also said the company will increasingly integrate reinforcement learning into its robots to broaden potential applications. 

"With reinforcement learning, we can use fewer policies to achieve greater terrain adaptation and coverage, making the robot more robust and intelligent," Ma said.