In recent years, the field of robotic portrait drawing has garnered considerable interest, as evidenced by the growing number of researchers focusing on either the speed or quality of the output drawing. However, the pursuit of either speed or quality alone has resulted in a trade-off between the two objectives. Therefore, in this paper, we propose a new approach that combines both objectives by leveraging advanced machine learning techniques and a variable line width Chinese calligraphy pen. Our proposed system emulates the human drawing process, which entails planning the sketch and creating it on the canvas, thus providing a realistic and high-quality output. One of the main challenges in portrait drawing is preserving the facial features, such as the eyes, mouth, nose, and hair, which are crucial for capturing the essence of a person. To overcome this challenge, we employ CycleGAN, a powerful technique that retains important facial details while transferring the visualized sketch onto the canvas. Moreover, we introduce the Drawing Motion Generation and Robot Motion Control Modules to transfer the visualized sketch onto a physical canvas. These modules enable our system to create high-quality portraits within seconds, surpassing existing methods in terms of both time efficiency and detail quality. Our proposed system was evaluated through extensive real-life experiments and showcased at the RoboWorld 2022 exhibition. During the exhibition, our system drew portraits of more than 40 visitors, yielding a survey outcome with a satisfaction rate of 95%. This result indicates the effectiveness of our approach in creating high-quality portraits that are not only visually pleasing but also accurate.

Recently, a diverse range of robots with various functionalities have become a part of our daily lives. However, these robots either lack an arm or have less capable arms, mainly used for gestures. Another characteristic of the robots is that they are wheeled-type robots, restricting their operation to even surfaces. Several software platforms proposed in prior research have often focused on quadrupedal robots equipped with manipulators. However, many of these platforms lacked a comprehensive system combining perception, navigation, locomotion, and manipulation. This research introduces a software framework for clearing household objects with a quadrupedal robot. The proposed software framework utilizes the perception of the robot’s environment through sensor inputs and organizes household objects to their designated locations. The proposed framework was verified by experiments within a simulation environment resembling the conditions of the RoboCup@Home 2021-virtual competition involving variations in objects and poses, where outcomes demonstrate promising performance.

This paper presents a system in which mobile manipulators and transport agents cooperate to solve a multi-agent pickup and delivery (MAPD) problem. The primary objective is to allocate appropriate tasks to heterogeneous robots by considering their capabilities and states. Unlike previous studies that focused on homogeneous teams or assigned distinct roles to heterogeneous robots, this work emphasizes synergy through cooperative task execution. A key feature of the proposed system is that mobile manipulators behave differently depending on whether they are paired with a transport agent. Additionally, rather than generating a full trajectory from start to end, the system plans partial trajectories, allowing dynamic re-pairing of transport agents through an auction algorithm. After re-pairing, new starting nodes are defined, and the following trajectory is updated accordingly. The proposed system is validated through simulations, and its effectiveness is demonstrated by comparing it against a baseline system without dynamic pairing.

Large language models (LLMs) represent a significant advancement in integrating physical robots with AI-driven systems. We showcase the capabilities of our framework within the context of the real-world household competition. This research introduces a framework that utilizes RDMM (Robotics Decision-Making Models), which possess the capacity for decision-making within domain-specific contexts, as well as an awareness of their personal knowledge and capabilities. The framework leverages information to enhance the autonomous decision-making of the system. In contrast to other approaches, our focus is on real-time, on-device solutions, successfully operating on hardware with as little as 8GB of memory. Our framework incorporates visual perception models equipping robots with understanding of their environment. Additionally, the framework has integrated real-time speech recognition capabilities, thus enhancing the human-robot interaction experience. Experimental results demonstrate that the RDMM framework can plan with an 93\% accuracy. Furthermore, we introduce a new dataset consisting of 27k planning instances, as well as 1.3k text-image annotated samples derived from the competition. The framework, benchmarks, datasets, and models developed in this work are publicly available on our GitHub repository at https://github.com/shadynasrat/RDMM.