We propose a visual servoing framework for learning to improve grasps of objects. RGB and depth images from grasp attempts are collected using an automated data collection process. The data is then used to train a Grasp Quality Network (GQN) that predicts the outcome of grasps from visual information. A grasp optimization pipeline uses homography models with the trained network to optimize the grasp success rate. We evaluate and compare several algorithms for adjusting the current gripper pose based on the current observation from a gripper-mounted camera to perform visual servoing. Evaluations in both simulated and hardware environments show considerable improvement in grasp robustness with models trained using less than 30K grasp trials. Success rates for grasping novel objects unseen during training increased from 18.5% to 81.0% in simulation, and from 17.8% to 78.0% in the real world.

In this paper we provide empirical evidence that using humanlike gaze cues during human-robot handovers can improve the timing and perceived quality of the handover event. Handovers serve as the foundation of many human-robot tasks. Fluent, legible handover interactions require appropriate nonverbal cues to signal handover intent, location and timing. Inspired by observations of human-human handovers, we implemented gaze behaviors on a PR2 humanoid robot. The robot handed over water bottles to a total of 102 naïve subjects while varying its gaze behaviour: no gaze, gaze designed to elicit shared attention at the handover location, and the shared attention gaze complemented with a turn-taking cue. We compared subject perception of and reaction time to the robot-initiated handovers across the three gaze conditions. Results indicate that subjects reach for the offered object significantly earlier when a robot provides a shared attention gaze cue during a handover. We also observed a statistical trend of subjects preferring handovers with turn-taking gaze cues over the other conditions. Our work demonstrates that gaze can play a key role in improving user experience of human-robot handovers, and help make handovers fast and fluent.