2D and 3D Grasp Planners for the GET Asymmetrical Gripper
New planner lifts success rates by 40% in just 683 milliseconds...
A team led by Andrew Goldberg, Edward H. Adelson, and Ken Goldberg has released two new grasp planners for the GET asymmetrical gripper: GET-2D-1.0 and GET-3D-1.0. The 2D version operates from a single-view RGB-D image, using the Ferrari-Canny metric and a novel sampling strategy to compute grasps in just 683 milliseconds. In physical experiments, it improved lift success, shake survival, and force resistance by over 40% compared to a bounding-box baseline.
The 3D variant, GET-3D-1.0, uses a mesh-based approach with ray-tracing and a full 3D gripper model. While it showed slight improvements in lift success and shake survival over the 2D planner, it is far more computationally expensive, averaging 17 seconds per planning cycle. The researchers suggest that for most practical applications, the speed and robustness of GET-2D-1.0 make it the preferred choice.
- GET-2D-1.0 processes single-view RGB-D images in 683 ms using the Ferrari-Canny metric
- Physical experiments show over 40% improvement in lift success, shake survival, and force resistance over baselines
- GET-3D-1.0 offers marginal gains but takes 17 seconds per plan, 25x slower than the 2D version
Why It Matters
Fast, robust grasp planning for asymmetrical grippers enables more reliable robotic manipulation in cluttered environments.