Franka Emika
7-DoF robotic arm used as primary simulation platform for LIBERO-Spatial and Robosuite Lift benchmarks
“Real-robot experiments are conducted on a Franka Panda fixed-arm manipulator with a single exocentric RGBD camera”
Source→“The pretraining corpus, NeoData, spans multiple robot platforms (ARX X5, UR5e, Flexiv, Franka, Piper) and a handheld UMI-style gripper, with every gripper finger carrying a self-developed visuo-tactile sensor”
Source→“The method generalizes to robot embodiments never seen during training. The model was finetuned on single-arm Franka Panda data from DROID, yet handles a bimanual R1-Pro humanoid robot from BEHAVIOR-1K. On the unseen bimanual embodiment, the method achieves PSNR of 22.90 vs. 18.39 for Ctrl-World”
Source→“The Franka Panda manipulator is used for real-world data collection with a custom 3D-printed compliant end-effector. The DROID dataset, which is central to training, is also primarily Franka-based”
Source→“The paper demonstrates that even with different PD gain settings in simulation, SILO allows zero-shot deployment without manual tuning”
Source→“The real world setup consists of a single robot arm (Franka Research 3 in our experiments)”
Source→“achieve a 73% success rate on six complex tasks using a 7-DoF arm and 16-DoF hand”
Source→“The simulation experiments used a '7DoF Franka Panda in the LIBERO environment' (Section 4). Franka remains the standard robot for simulation benchmarks in manipulation research.”
Source→“Robot manipulator used for all hardware experiments. Standard hardware platform; results are demonstrated on a widely-deployed arm, increasing reproducibility relevance.”
Source→“The Euler-Lagrange dynamics equations in Branch B are specifically derived for the Franka Panda's kinematics.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.