Towards human-like handover timing performance with legged manipulators

Deploying perception modules for human-robot handovers is challenging because they require a high degree of reactivity, generalizability, and robustness to work reliably for a diversity of objects. Further complications arise as each object can be handed over in a variety of ways, that can cause occlusions and viewpoint changes. On legged robots, deployment is particularly challenging because of the limited computational resources and the additional image-space noise resulting from locomotion. In this paper, we introduce an efficient and object-agnostic real-time tracking framework, specifically designed for handover tasks between a human and a legged manipulator. The proposed method combines fast optical flow with Siamese-network-based tracking and depth segmentation in an adaptive Kalman Filter framework. We show that we outperform the state-of-the-art for tracking during human-robot handovers with our legged manipulator system. We demonstrate the generalizability, reactivity, and robustness of our system through experiments in different handover scenarios and by carrying out a user study. Furthermore, as timing has been proven to be more important than spatial accuracy in human-robot interaction tasks, we show that we reach close to human timing performance not only in terms of objective metrics, such as handover and reaction time but also by considering subjective metrics gathered from the participants in the user study.