Global Academic Insights: Elephant Robotics Unveils Groundbreaking Research Collection on Cobots, AGVs and Compound Mobile Robots from Top Universitie

In the ever-evolving landscape of robotics, the symbiosis between humans and machines takes center stage, marking the advent of a new era of innovation. With the development of robotics technology, the application fields of collaborative robots and compound mobile robots are becoming more and more abundant. Today, Elephant Robotics is thrilled to introduce a pioneering collection of academic papers focusing on Collaborative Robots (Cobots), Automated Guided Vehicles (AGVs) and compound mobile robots sourced from prestigious global universities, such as Carnegie Mellon University, University of California, Berkeley, Imperial College London, University of Bristol, The Hong Kong University of Science and Technology and Waseda University. This compilation delves into the transformative capabilities of Elephant Robotics' 6 DOF collaborative robots, myCobot series, mobile robot, myAGV and autonomous compound mobile robot, LIMO Cobot, highlighting the innovative applications of robotics technology in different fields.

From over 100 pioneering academic papers focusing on Elephant Robotics' robotic products from top global institutions, we have meticulously selected 15 notable studies. This comprehensive collection illustrates how robotics technology is driving advancements across various sectors, including industry, healthcare, agriculture, construction, and manufacturing. It serves as a valuable knowledge repository, showcasing significant theoretical breakthroughs alongside significant practical applications, and providing a holistic perspective on the future of human-robot interaction and collaboration. Tailored for enthusiasts, professionals, and researchers alike, this collection is an essential resource for anyone looking to stay informed about the latest developments and innovations in the robotics field and industries.

1. Topic: FogROS2-LS: A Location-Independent Fog Robotics Framework for Latency Sensitive ROS2 Applications

Authors: Kaiyuan Chen, Michael Wang, Marcus Gualtieri, Nan Tian, Christian Juette, Liu Ren, Jeffrey Ichnowski, John Kubiatowicz and Ken Goldberg

Universities: University of California, Berkeley, Carnegie Mellon University

Abstract: This study presents FogROS2-LS, an independent Fog Robotics framework designed specifically for delay-sensitive ROS2 applications. The framework addresses the latency issues in cloud robotics caused by network fluctuations by extending anycast routing, thereby establishing secure and low-latency connections between robots and cloud servers. It enables seamless migration of state estimators and controllers to cloud and edge devices without requiring modifications to existing ROS2 applications. FogROS2-LS dynamically selects optimal service deployment to meet latency requirements, allowing resource-constrained robots to navigate safely in complex environments. The project utilizes Elephant Robotics' mobile robot myAGV to validate the effectiveness of the FogROS2-LS framework through collision avoidance and target tracking scenarios, showcasing its capabilities in network fault recovery and continuous target tracking. These experiments showcase FogROS2-LS's potential to redefine latency-sensitive robotic operations.

2. Topic: Toward Internet of Human and Intelligent Robotic Things With a Digital Twin-Based Mixed Reality Framework

Authors: Dandan Zhang, Ziniu Wu, Jin Zheng, Yifan Li, Zheng Dong, and Jialin Lin

Universities: Imperial College London, University of Bristol

Abstract: This article presents the HuBotVerse framework, designed to facilitate the integration of humans and Intelligent Robot Things (IoHIRT). The framework emphasizes security, user-friendliness, manageability, and open-source accessibility, allowing for the integration of various human-machine interaction interfaces to enhance collaborative control. It seamlessly incorporates a range of Human-Robot Interaction (HRI) interfaces, promoting effective collaboration between humans and robots. Utilizing a digital twin (DT) mixed reality (MR) interface, HuBotVerse provides users with intuitive and immersive interaction methods, significantly improving the efficiency of remote operations. The research conducted with the 6 DOF collaborative robot arm, myCobot 320, validates the development of the IoHIRT concept and the HuBotVerse framework. The integration of digital twin (DT)-based mixed reality (MR) for ergonomic interaction positions HuBotVerse as a promising solution for human-robot collaborative control. This innovative framework enhances both teleoperational efficiency and user experience, offering notable advantages in applications related to homecare services and healthcare.

3. Topic: A Novel Teleoperation Approach Based on MediaPipe and LSTM

Authors: Jianan Xie, Zhen Xu, Jiayu Zeng, Xiaohan Du, Yilin Zhang, Shanshan Wang, Hongming Chen, and Kenji Hashimoto

University: Waseda University

Abstract: The article introduces a novel remote control method for operating mobile robots, enabling operators to control compound mobile robots with one hand. The research team combines MediaPipe Hands hand skeletal detection technology with RGB-D cameras to capture 3D keypoint coordinates more accurately. By analyzing various spatial features of the hand, corresponding remote control commands are generated. To facilitate switching between control objects using specific gestures, a gesture recognition model based on Long Short-Term Memory (LSTM) architecture achieves 100% accuracy in recognizing three distinct gestures. This project utilizes the autonomous compound mobile robot, LIMO Cobot, as a testbed, incorporating a custom inverse kinematics (IK) solver to map the hand's position coordinates to the working space of the 6 DOF robotic arm, myCobot 280 M5. The IK solver calculates the joint angles required for the robotic arm to move to the designated position and activate a doorbell, effectively validating the proposed gesture-based remote control method.

4. Topic: One-Handed Wonders: A Remote Control Method Based on Hand Gesture for Mobile Manipulator

Authors: Jianan Xie, Yilin Zhang, Zhen Xu, Yuyang Gao, Jiawei Bai, Jiayu Zeng, and Kenji Hashimoto

University: Waseda University

Abstract: The Authorss have developed a one-handed gesture-based remote control method for operating mobile robots, enabling operators to manage an entire compound mobile robot with just one hand. Utilizing the autonomous compound mobile robot, LIMO Cobot, as the experimental platform, the project integrates MediaPipe's real-time hand keypoint detection technology with the RealSense D435i depth camera, addressing issues of depth recognition inaccuracies found in previous methods. By analyzing the hand's position, pitch, and rotation, corresponding control commands are generated. To facilitate switching between control objects using specific gestures, the research team introduced a lightweight gesture recognition model based on Gated Recurrent Units (GRU), achieving 100% accuracy in recognizing three distinct gestures. Finally, physical experiments conducted on the LIMO Cobot within a mobile manipulator operation platform provided preliminary validation of the proposed method's effectiveness. This research marks a milestone in intuitive and efficient human-robot interaction, especially in scenarios where minimal physical involvement from the operator is required.

5. Topic: AR-enhanced digital twin for human–robot interaction in manufacturing systems

Authors: Zhongyuan Liao, Yi Cai

University: The Hong Kong University of Science and Technology

Abstract: This paper presents a system that integrates Augmented Reality (AR) technology with Digital Twin (DT) capabilities to enhance human-robot interaction (HRI) in manufacturing environments. The system is designed to operate at 3 distinct levels of DT functionality: the virtual twin for in-situ monitoring, the hybrid twin for intuitive interaction, and the cognitive twin for optimized operation. Utilizing the 6 DOF robot arm, myCobot 280 Pi, as the experimental platform, user studies validated the effectiveness of the AR-enhanced DT system in reducing operation time, lowering error rates, and improving the overall user experience. This innovative system offers a new solution for the future of smart manufacturing by combining AR and DT technologies, thereby enhancing the intuitiveness and efficiency of robotic operations.

6. Topic: TrojanRobot: Backdoor Attacks Against Robotic Manipulation in the Physical World

Authors: Xianlong Wang, Hewen Pan, Hangtao Zhang, Minghui Li, Shengshan Hu, Ziqi Zhou, Lulu Xue, Peijin Guo, Yichen Wang, Wei Wan, Aishan Liu, and Leo Yu Zhang

Universities: Huazhong University of Science and Technology, Beihang University, Griffith University

Abstract: This paper explores the issue of backdoor attacks faced by robots in the physical world and presents a novel approach to such attacks for the first time. Researchers embedded a backdoor large vision-language models (LVLMs) into the visual perception module of the robotic system, effectively misdirecting operations of the 6 DOF collaborative robotic arm myCobot 280 Pi through the use of everyday items as triggers. The experiments demonstrate that this backdoor attack strategy can effectively misguide the robotic arm’s actions when common objects are employed as triggers. The study makes several key contributions. It is the first to investigate existing robotic manipulation schemes, revealing that traditional backdoor attacks are difficult to adapt directly to robotic systems. Additionally, it introduces a plugin-based backdoor model that modifies inputs to the visual perception module, facilitating effective and covert attacks. Finally, it provides the first experimental validation of the proposed robotic backdoor attack’s efficacy in physical world application scenarios.

7. Topic: BestMan: A Modular Mobile Manipulator Platform for Embodied AI with Unified Simulation-Hardware APIs

Authors: Kui Yang, Nieqing Cao, Yan Ding, and Chao Chen

Universities: Xi'an Jiaotong University, Chongqing University

Abstract: This paper presents the BestMan platform, a modular mobile manipulator platform designed for research in Embodied Artificial Intelligence (Embodied AI). Built on the PyBullet simulator, the platform offers a unified simulation and hardware API. This approach effectively addresses the complexities of multi-level technology integration, the modularity limitations of existing platforms, and the challenges of interfacing between simulation environments and physical robotic systems. BestMan enables a wide range of service tasks for robots in home setting by offering 10 key components, each targeting specific areas of Embodied AI research and facilitating algorithm implementation and customization. Notably, the 6 DOF collaborative robot arm, myCobot Pro 630, is a key hardware component of the platform, showcasing its adaptability and software decoupling capabilities. Within the BestMan framework, myCobot Pro 630 can seamlessly integrate with other components to perform diverse tasks, including 3D vision-guided classification and grasping, applications involving compound mobile robots in conjunction with AGVs, and G-code guided artistic creations, showcasing its practical value in Embodied AI research.

8. Topic: Enhancement of Control Performance for Degraded Robot Manipulators Using Digital Twin and Proximal Policy Optimization

Authors: SU-YOUNG PARK, CHEONGHWA LEE, HYUNGJUNG KIM, AND SUNG-HOON AHN

Universities: Seoul National University, Konkuk University

Abstract: The study introduces a pioneering method to enhance the control performance of degraded robot manipulators. Utilizing digital twins and proximal policy optimization (PPO), a deep reinforcement learning algorithm, this novel approach tackles the costly and time-consuming challenges of repairing or replacing sophisticated robot hardware and control systems. By simulating an unstable 6 DOF robotic arm, the research utilizes domain randomization and deep learning techniques to enhance the accuracy and stability of target positioning. Experimental results demonstrate that this approach significantly reduces positional errors, outperforming traditional control methods. Using the 6 DOF desktop robotic arm, myCobot 280 Pi, as the experimental platform, the study randomizes digital twin parameters to simulate performance degradation and develop more robust control strategies. These strategies are applicable to various multi-axis collaborative robots, providing a solution to maintain performance while reducing costs.

9. Topic: An Approach to Elicit Human-Understandable Robot Expressions to Support Human-Robot Interaction

Authors: Jan Leusmann, Steeven Villa, Thomas Liang, Chao Wang, Albrecht Schmidt, and Sven Mayer

University: Ludwig Maximilian University of Munich

Abstract: This paper introduces a two-stage method for developing and validating robot expressions that are understandable to humans, aimed at enhancing the naturalness and intuitiveness of human-robot interaction. In the first stage, the approach utilizes human imitation and movement to generate expressive behaviors. In the second stage, user studies are conducted to validate the comprehensibility of these expressions. The project utilizes the world's smallest and lightest 6 DOF robotic arm, myCobot 280, as a case study, demonstrating how to design robotic movements that express curiosity and attention. Experimental results confirm the effectiveness of these expressions. This method contributes a systematic approach to generating and validating non-verbal communication in robots, ultimately improving communication efficiency and interaction quality between humans and machines. It marks a significant advancement in the field of human-robot interaction (HRI).

10. Topic: 3D Data Collection for Individual Plant Farming

Authors: Jacob Karty, Blake Hament

University: Elon University

Abstract: The paper introduces an automated phenotyping analysis technique that tracks plant health and lifecycle to promote water conservation, reduce pesticide reliance, and increase yields. Utilizing the 6 DOF collaborative robotic arm, myCobot 320 M5, equipped with an RGB camera, the system captures 2D data of plants from multiple angles to create 3D models. The project addresses the workspace limitations of the robotic arm by integrating inverse kinematics and photometric reconstruction techniques to stitch images into a point cloud. Electronic simulations are used to determine optimal shooting angles for maximum plant coverage. Experimental results validate the effectiveness of this method in data collection for photometric reconstruction, enhancing workspace efficiency and providing a robust dataset for machine learning in precision agriculture to extract phenotypic information from plants.

11. Topic: Development of a Retrofit Backhoe Teleoperation System using Cat Command

Authors: Koshi Shibata, Yuki Nishiura, Yusuke Tamaishi, Kohei Matsumoto, Kazuto Nakashima, and Ryo Kurazume

University: Kyushu University

Abstract: This paper presents a novel backhoe remote operation system that is easy to install and cost-effective. The system integrates remote control and sensing technologies, retrofitting the 6 DOF robotic arm, myCobot 280 M5, and electric cylinders onto the Cat Command remote control system, enabling remote operation via a 5G network. Importantly, the system requires no modifications to the cab control unit, allowing operators to remain inside the cab while controlling the machinery remotely. By swapping connection components, the system can manage various devices. The sensing system utilizes sensors and a 360° camera to transmit real-time images from the field, resulting in a more compact and lightweight setup compared to traditional equipment. The project successfully validated the system's effectiveness, confirming its suitability for practical soil excavation tasks. This development marks a significant advancement in the field of remote construction equipment operation, promising increased efficiency and safety in hazardous construction environments.

12. Topic: Development of an Intuitive Mixed Reality Human Robot Interaction Interface for Construction Applications

Authors: Tennakoon*, Jadidi, M.*, and RazaviAlavi S.R.

University: York University

Abstract: This research addresses low productivity and worker safety issues in the construction industry by developing a hybrid reality (MR) human-robot interaction (HRI) interface that combines the precision of robotics with the on-site experience of workers. By installing a ZEDM camera at the end of the 6 DOF collaborative robot, myCobot 320 M5, and utilizing the Meta Quest 2 VR headset to provide a first-person perspective, the project employs gesture recognition technology to enable comprehensive control of the robot. The results demonstrate that this MR HRI interface enhances depth perception and situational awareness, leading to improved operational efficiency. This interface promotes the application of robotics in the construction sector, allowing workers to safely perform hazardous tasks remotely, thereby validating the potential of mixed reality human-robot interaction technologies in the industry.

13. Topic: Six Degrees of Planning: Automated Planning for Surgical Navigation Under MyCobot’s Six Degrees of Freedom

Authors: Colton Barr*, Mateus Karvat Camara*, Sidney Givigi

University: Queen's University

Abstract: This paper investigates the application of medical robotics in stereoelectroencephalography (SEEG) surgery, focusing on how automated planning techniques can control the 6DOF collaborative robot arm, myCobot 280. Researchers employed hybrid planning and the Expression-based Numerical Heuristic Search Planner (ENHSP) to simulate robotic joint movements, achieving precise needle placement at 3D coordinates within the brain. The study evaluated the planner's performance across various planning domains and visualized results using 3D Slicer simulations. Findings indicate that this method effectively prevents collisions with the patient's head, generating joint rotation movements to accurately position the needle. The research also outlines future directions, including simulating link collisions, enabling simultaneous multi-joint movements, and developing customized heuristic methods to optimize planner performance.

14. Topic: Gesture-Controlled Robotic Arm for Agricultural Harvesting Using a Data Glove with Bending Sensor and OptiTrack Systems

Authors: Zeping Yu, Chenghong Lu, Yunhao Zhang and Lei Jing*

University: University of Aizu

Abstract: This paper introduces a gesture-controlled robotic arm system designed for agricultural harvesting, utilizing data gloves to capture hand movements and gestures for precise control of the 6 DOF collaborative robotic arm, myCobot 320 Pi. The system employs bending sensors and OptiTrack spatial tracking technology, combined with a CNN+BiLSTM machine learning model, to accurately recognize gestures for robotic arm manipulation. Experiments demonstrate high precision in replicating hand movements, with an Euclidean distance of 0.0131 meters and a root mean square error (RMSE) of 0.0095 meters, achieving a recognition accuracy of 96.43%. By integrating the advantages of both semi-automated and fully automated systems, this innovative solution offers an effective approach to enhance efficiency in agricultural harvesting and reduce the labor-intensive nature of fruit picking.

15. Topic: Research & analysis of automated robots using sensors and path finding algorithms

University: Uppsala University

Abstract: This paper's thesis investigates the concept of robotic automation, along with the relevant literature and technical requirements, aiming to design a small-scale robotic demonstrator capable of autonomous operation. The research utilizes the mobile robot myAGV for environmental navigation and the collaborative robot myCobot 280 for object interaction. The focus of the study is on employing optical detection and LiDAR sensors to collect data and generate maps, while utilizing path planning algorithms to reach targets within the map while avoiding walls and obstacles. The results showcase the capabilities of the compound mobile robot in object interaction, autonomous navigation, data collection, and environmental mapping, as well as performance analysis of the algorithms through simulations. This development not only enhances productivity and safety but also underscores the ethical implications of integrating autonomous systems into real-world applications, setting a new benchmark for the future of robotics in dynamic and sensitive settings.

This curated collection of academic papers delves into groundbreaking research focused on innovative design tools for close-proximity human-robot collaboration. Analyzing these cutting-edge studies reveals the immense potential of robotics technology to enhance efficiency, reduce costs, improve safety, and elevate user experience. The applications of robotics are rapidly expanding, from homecare to precision agriculture, and from construction to medical surgery, making them an indispensable part of our daily lives. As technology continues to evolve, we believe robotics will bring even more innovation and convenience into everyday life. Elephant Robotics is committed to ongoing innovation, providing advanced robotics solutions to meet evolving market demands while actively expanding the scope of robotic application scenarios.

To view and download selected academic papers from 2024, please click the following link: https://drive.google.com/drive/folders/12euaZG-alvIGkQVRimv8fNxd8VYv1V2j?usp=sharing.