Rather than equip their sharp-looking general-purpose GR-1 humanoid with a full suite of next-generation sensors, including things like radar and LiDAR, Fourier Intelligence’s engineers opted for vision only.
The GR-1 has six RGB cameras around its frame for a 360-degree view of the world around it. This setup also provides a bird’s eye view map using camera data and a neural network that learns from context to create 3D spatial features and virtual objects.
The company says the technology “then converts the data into a three-dimensional occupancy grid, helping the GR-1 navigate traversable and impassable areas.” The bot recently conducted outdoor walking tests where it is said to have demonstrated “high efficiency and accuracy in detecting vehicles and pedestrians along sidewalks” in real time.
Fourier Camera-Only Perception Network: BEV+Transformer+Occupancy
As with Tesla’s decision in 2021 to operate its Autopilot systems using mostly vision systems, this development path should significantly reduce hardware costs – all while “enhancing the GR-1’s perception of the environment, achieving safer and more efficient operations with precision as in humans”.
The current model of the GR-1 looks very different from the skeletal, open-faced bipod prototypes we introduced last year. Fourier’s product page says it’s capable of walking at human walking speed on a variety of surfaces, with adaptive balance algorithms to help keep it upright when going up or down slopes.
The company says it has 54 degrees of freedom on the body – which means three in the head and waist, seven in each arm, and eleven are each five-digit hand and six in each leg. The bot boasts a maximum torque of 230Nm and also has funky purple side bumpers – why can’t a humanoid be a slave to fashion?
Fourier intelligence
The intelligent AI is said to include “a multi-modal language model similar to ChatGPT as well as advanced semantic knowledge, natural language processing and logical reasoning”.
The once empty head now has a face that is home to a high-definition display, audio speakers and a microphone. The omni vision system not only allows him to map and navigate his surroundings in real time, but also supports his abilities to avoid obstacles and collisions.
“This advance represents a new stage in our research in the field of artificial intelligence,” said Roger Cai, the company’s director of robotics application research and development. “With our clear vision solution, the GR-1 is poised to play a key role in a variety of applications such as medical rehabilitation, family services, reception and counseling, security inspection, emergency response and industrial production.”
Source: Fourier Intelligence