In current robotic arm grasping objects in real-world scenarios, cameras are typically installed on the arm to determine the object's location through image recognition, thereby adjusting the arm's operating position. However, the greatest challenge lies in controlling the force with which the arm grasps the object. NVIDIA's proposed 6-DoF GraspNet computing framework enables robotic arms to more efficiently learn how to grasp objects anywhere in space.
The 6-DoF GraspNet computing framework allows the robotic arm to determine the X, Y, and Z coordinates (left-right, front-back, and height) in space through image recognition, and incorporates rotational motion to generate data for a total of six axes. This allows for faster identification of objects in different spatial locations, avoiding the need for the robotic arm to relearn and re-identify objects after their positions change, which would slow down its work efficiency.
According to NVIDIA, under this computing framework operating mode, the robotic arm can determine the location of an object through real-time vision. The FleX evaluation mode based on particle simulation computing technology allows the robotic arm to make judgments based on actual measurement data, allowing it to quickly determine the actual location of an object even without prior deep learning training.
On the other hand, the gripping evaluation mode is also used for the robot arm's grasping of objects. This allows the robot arm to be tested through different force levels to evaluate the actual force required to grasp the object, thereby successfully grasping the object and avoiding crushing the object due to excessive force.
This technology can be applied in different scenarios through modularization and can even be used in conjunction with other computing frameworks, allowing robotic arms to complete tasks with more precise movements.
Such applications not only enable robotic arms to operate more precisely, but also reduce the problem of improper damage to goods when robotic arms are used in applications such as cargo sorting, thereby making applications such as unmanned stores and smart warehousing more popular.
