Researchers at the University of California, San Diego have successfully developed a humanoid robot capable of learning and executing numerous expressive actions, including simple dance routines and gestures such as waving, high-fiving, and hugging, while maintaining a steady pace on various terrain surfaces.
The enhanced expressiveness and dexterity of these humanoid robots create a pathway for enhancing human-robot interactions in settings such as manufacturing facilities, hospitals, and homes, where robots can operate safely alongside humans and potentially replace them in hazardous environments like laboratories or disaster sites?
“Through human-like physical expressions, our objective is to build trust and demonstrate the feasibility of robots living harmoniously alongside humans,” said Xiaolong Wang, a professor in the Department of Electrical and Computer Engineering at the University of California, San Diego’s Jacobs School of Engineering. “We’re committed to repositioning the public’s perception of robots as friendly and cooperative, rather than fear-inducing like The Terminator.”
Wang and his team will be presenting their research at the 2024 Robotics: Science and Techniques Convention, scheduled to take place from July 15 to 19 in Delft, Netherlands.
What sets this humanoid robot apart is its remarkable ability to execute a diverse range of human physical movements, allowing it to learn and replicate novel actions with remarkable facility. With an ease comparable to that of a nimble dance scholar, robots can rapidly absorb and replicate novel movements.
To train their robots, the group employed a comprehensive array of motion capture data and dance videos. Their approach focused on personalizing training for both the upper and lower body separately. The robotic’s taller stature enabled it to replicate various reference movements, such as dancing and high-fiving, while its legs focused on a subtle stepping motion to maintain balance and traverse diverse terrain.
“At the core, our goal is to demonstrate the adaptability of the robot in performing diverse tasks as it navigates through space without stumbling,” Wang explained.
Despite individualized coaching for both upper and lower body, the robot functions under a cohesive policy governing its overall architecture. This comprehensive coverage enables the robot to execute sophisticated physical movements while walking steadily across diverse surfaces, including gravel, dirt, wood chips, grass, and sloping concrete pathways.
Following initial simulations on a digital humanoid robot, the concept was successfully tested on a physical prototype. The robot demonstrated remarkable flexibility, seamlessly executing both familiar and novel actions in real-world scenarios with ease.
Currently, the robot’s movements are controlled by a human operator using a game controller, determining its speed, trajectory, and specific gestures. The team envisions a cutting-edge robotic prototype equipped with a high-resolution digital camera, enabling it to perform tasks and traverse complex environments independently, leveraging advanced autonomous navigation capabilities.
The team is focusing on refining the robotic’s design to tackle increasingly complex and minute tasks. “With enhanced physical capabilities, we can broaden the scope of movements and actions that our robotic system is capable of performing,” stated Wang.
“Entire-Physique Expressive Control for Humanoid Robotics: A Collaborative Effort”
The authors share equal credit for their contributions to this project.
