Pick and place - Research demonstration - Pick objects from a moving conveyor and sort them by colour in physical learned-policy trials. - environment: Laboratory conveyor pick-and-sort setup - Hardware: Right Rizon 4S with Wuji Hand 2; Second static Rizon 4S supplies an additional camera view; Wrist fisheye cameras and head stereo camera - Software: Learned vision-action policy; teleoperated training demonstrations - Control: autonomous - source: arxiv.org
Precision insertion - Research demonstration - Insert 3D-printed pegs into matching holes using a simulated-training policy driven by joint and gripper state. - environment: Laboratory peg-in-hole fixtures - Hardware: Rizon 4S with Robotiq 2F-85; 3D-printed peg and hole fixtures; No camera, F/T input or hole pose enters the policy - Software: Frozen sim-to-real policy at 50 Hz - Control: autonomous - source: arxiv.org
Force-controlled manipulation - Research demonstration - Slide an end effector along a force scale while targeting 10 N normal force in physical policy evaluations. - environment: Laboratory contact-force test setup - Hardware: Rizon 4S with six-axis wrist F/T sensor; Custom compliant fingertips and task-specific joystick hardware; Force scale - Software: ForceWAM learned controller using demonstration and play data - Control: autonomous - source: arxiv.org