I. INTRODUCTION

There are many situations and environments where we need robots to replace humans at the site. Despite the recent progress in robot cognition based on AI techniques, fully autonomous solutions are still far from producing socially and physically competent robot behaviors; that is why teleoperating robots (Fig. 1) acting as physical avatars of human workers at the site is the most reasonable solution. In environments like construction sites, chemical plants, contaminated areas and space, teleoperated robots could be extremely valuable, relieving humans from any potential hazard. Contrary to other conventional robotic platforms, humanoids’ structure is a better fit for environments and tasks that are designed for and performed by humans. The operational versatility of these robots makes them suitable for work activities that require a variety of complex mobility and manipulation skills, such as inspection, maintenance, and interaction with human operators. In certain contexts such as telenursing, where human subjects are expected to interact with a teleoperated robot, the human-likeness factor is important since it increases the acceptability, social closeness, and legibility of its intentions .

Fig 1: Examples of humanoid robot teleoperation; from top left to bottom right corner: , , , , , , , .

In the literature, different attempts have been made to

deploy the senses, actions, and presence of a human to a remote location in real-time, leading to a more connected world .

Inspired by a visit to the Tachi Lab, the XPRIZE Foundation has recently launched the ANA Avatar XPRIZE global competition . Previously, in response to the 2011 Fukushima Daiichi nuclear disaster, the DARPA Robotics Challenge (DRC) was launched to promote innovation in human-supervised robotic technology. In space applications, rovers and mobile manipulators were teleoperated from aboard the International Space Station (ISS), in the context of METERON and Kontur-2 projects . In 2019, the humanoid Skybot F-850 was rocketed to the ISS ; however, it turned out to have a design that did not work well, demonstrating that there is still work to do to get humanoids into space.

Humanoid robot teleoperation involves many multidisciplinary and interleaved challenges, ranging from dynamics and control to communication and human psychophysiology. Uniquely, due to their resemblance to human appearance, societal expectations are high as well; they are expected to do a wide range of tasks that are not expected from other types of robots. They are highly redundant with nonlinear, hybrid, and underactuated models. While doing dynamic and agile motions with the feet like walking, running, or stepping over obstacles, they are supposed to perform dexterous power and precision manipulation. At the same time, they are expected to work alongside humans, be safe, friendly, and socially interact with others. On the other hand, teleoperation interfaces and techniques should be designed such that the human operator receives minimal, effective, and informative haptic feedback from the humanoid robot, to cover for human errors, overcome communication delays, and above all, be telepresent. Along with these challenges, the field is new and due to its high resource demand for development, not many laboratories have been working on it.

Many efforts from the robotics community have been devoted to studying humanoid robots, teleoperation, evaluation metrics, or human-robot interaction. Among them, the book on humanoid robotics studied comprehensively different aspects of humanoid robots, including their history, design, mechanics, control, simulation, and interaction. Several survey papers likewise studied specific aspects of humanoid robots, for example humanoid dynamics , control , motion generation , or robot teleoperation interface design and metrics . A primary work on bilateral teleoperation techniques has been presented by as well. Another seminal survey covers many aspects of interactive robots, including their design, autonomy level, and human factors that helped us in articulating the current manuscript. Another interesting survey highlights several aspects of humanoid teleoperation and autonomy. However, is a decade old, and an up-to-date survey on the topic is missing, especially considering that humanoid teleoperation is a far-from-solved challenge and a highly active field of research where new solutions are proposed each year. Following the workshop in , this survey paper presents the latest results in humanoid robot teleoperation and draws in detail the challenges that the research community faces to effectively deploy such systems toward real-world scenarios.

Starting from what emerged from the workshop, we conducted a survey on teleoperation of humanoid robots. We present here the systems and devices that have been adopted so far to teleoperate humanoids (Sec. II) and how these robots have been modeled, retargeted, and controlled (Sec. III). We also examine a promising case of teleoperation in which the robot assists the user in accomplishing a desired task (Sec. IV). Later, we discuss complications along with some compensating solutions that arise due to non-ideal communication channels (Sec. V). We explain the evaluation of teleoperation systems prior to development to meet the users’ needs (Sec. VI). Finally, discussions on current and potential applications and the associated challenges follow (Sec. VII).