Researchers studied how reinforcement learning enabled a robot arm equipped with a parallel jaw gripper and tactile sensors to insert unknown objects into a hole.