Learning by Watching: A Review of Video-based Learning Approaches for Robot Manipulation

Robot learning of manipulation skills is hindered by the scarcity of diverse, unbiased datasets. While curated datasets can help, challenges remain in generalizability and real-world transfer. Meanwhile, large-scale "in-the-wild" video datasets have driven progress in computer vision through self-supervised techniques. Translating this to robotics, recent works have explored learning manipulation skills by passively watching abundant videos sourced online. Showing promising results, such video-based learning paradigms provide scalable supervision while reducing dataset bias. This survey reviews foundations such as video feature representation learning techniques, object affordance understanding, 3D hand/body modeling, and large-scale robot resources, as well as emerging techniques for acquiring robot manipulation skills from uncontrolled video demonstrations. We discuss how learning only from observing large-scale human videos can enhance generalization and sample efficiency for robotic manipulation. The survey summarizes video-based learning approaches, analyses their benefits over standard datasets, survey metrics, and benchmarks, and discusses open challenges and future directions in this nascent domain at the intersection of computer vision, natural language processing, and robot learning.

Internet of Things Meets Robotics: A Survey of Cloud-based Robots

This work presents a survey of existing literature on the fusion of the Internet of Things (IoT) with robotics and explores the integration of these technologies for the development of the Internet of Robotics Things (IoRT). The survey focuses on the applications of IoRT in healthcare and agriculture, while also addressing key concerns regarding the adoption of IoT and robotics. Additionally, an online survey was conducted to examine how companies utilize IoT technology in their organizations. The findings highlight the benefits of IoT in improving customer experience, reducing costs, and accelerating product development. However, concerns regarding unauthorized access, data breaches, and privacy need to be addressed for successful IoT deployment.

Enhancing Human-robot Collaboration by Exploring Intuitive Augmented Reality Design Representations

As the use of Augmented Reality (AR) to enhance interactions between human agents and robotic systems in a work environment continues to grow, robots must communicate their intents in informative yet straightforward ways. This improves the human agent's feeling of trust and safety in the work environment while also reducing task completion time. To this end, we discuss a set of guidelines for the systematic design of AR interfaces for Human-Robot Interaction (HRI) systems. Furthermore, we develop design frameworks that would ride on these guidelines and serve as a base for researchers seeking to explore this direction further. We develop a series of designs for visually representing the robot's planned path and reactions, which we evaluate by conducting a user survey involving 14 participants. Subjects were given different design representations to review and rate based on their intuitiveness and informativeness. The collated results showed that our design representations significantly improved the participants' ease of understanding the robot's intents over the baselines for the robot's proposed navigation path, planned arm trajectory, and reactions.