Robotiq is a Canadian industrial robotics company that designs and manufactures adaptive grippers, vacuum grippers, force/torque sensors, wrist cameras, and palletizing workcells to help manufacturers automate repetitive tasks with collaborative robots. They also develop supporting software and pioneered "Lean Robotics," a methodology for deploying scalable, plug-and-play cobot automation across industries like machine tending, assembly, and palletizing.
“The real-world experiment used a Delto DG-3F-B gripper that was never in the training data, yet achieved 78% success by mapping Robotiq 3-Finger predictions with "constant offsets on certain joints and fixed values for the additional joints" (Section 5.3).”
Source→“A single trained model generates grasps for three morphologically distinct grippers — the Franka Panda (parallel-jaw), Robotiq 3-Finger (underactuated), and Allegro Hand (16-DOF dexterous) — by conditioning on gripper point cloud geometry and a learnable per-gripper query vector. The model achieves 71.38%, 72.16%, and 72.07% success rates respectively”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.