International Journal of Intelligent Robotics and Applications最新文献

A new soft pneumatic microactuator based on alternative pole water electrolysis has recently been proposed. In these actuators, a water-based electrolyte is electrolyzed under an alternative current, generating hydrogen/oxygen nanobubbles/microbubbles. These bubbles cause the expansion of the electrolyte, resulting in the displacement of the actuator membrane. These actuators stand out for their lightweight design, cost-effectiveness, high performance, and versatility for various applications. In this paper, a strong and fast millimeter-sized actuator based on alternative pole water electrolysis is proposed. The proposed actuator, electronic driver circuits, and measurement systems is implemented, and some experiments to investigate the actuator’s performance under different conditions, including input variables such as voltage, time, temperature, and mass load are conducted. Our experimental results and comparisons with other actuators demonstrate that the proposed actuator exhibits favorable properties in terms of response time, output mechanical force, reliability, scalability, and simplicity of manufacturing. The versatility of this actuator makes it suitable for a wide range of soft robotics applications, including limb movement and manipulation. Additionally, it has potential medical applications such as microrobotics for navigation in narrow body channels for diagnosis, sampling, drug delivery, and surgery.

Graphical abstract

Abstract

Quadrotors have been more frequently used in different areas, from aerial photography to drug delivery in medical emergencies. These vehicles have high maneuverability, which makes them suitable for carrying out missions that humans would not be able to do due to physical constraints. They can be used in inhospitable environments where the physical integrity and health of humans would be compromised. However, they are highly nonlinear and multivariable systems whose dynamics are strongly coupled. These characteristics turn attitude control design into a complex task. Furthermore, the controller has to be able to deal with uncertainties and exogenous disturbances in practice, intensifying the difficulty of the control problem. Therefore, a quadrotor attitude control must have high robustness and fast response without compromising its global stability. Aiming to gather solutions to this control problem, this article provides a detailed and in-depth discussion on quadrotor attitude control strategies for flight control designers, including angular representation, controller stability, fault tolerance, actuator saturation, and strategies for exogenous disturbance rejection.

Shallow Water Inspection & Monitoring Robot (SWIM-R) is designed to quickly and safely inspect oil and gas pipelines in extremely shallow waters. Divers clean and inspect pipeline joints. However, diving operations are slow in shallow waters as diving support ships cannot access shallow depths. Remotely operated vehicles (ROVs) that can perform cleaning and inspection are typically suited for deeper regions and are too large for smaller boats that navigate in shallow areas. To resolve this challenge, two SWIM-R vehicles and a companion Autonomous Surface Vehicle (ASV) were developed as a multi-robot system to minimize the reliance on divers for pipeline inspection. A unique mission architecture is presented that avails three operating modes depending on the depth; direct control from the shore, relayed control via the ASV, and direct control from a small zodiac. The mission architecture includes two ROVs; a Cleaning SWIM-R fitted with a water-jet nozzle to clean marine growth from the surface to be inspected, and an Inspection SWIM-R fitted with a neutrally-buoyant multi-functional robotic arm to inspect the surface and crawling tracks to traverse the seafloor. This multi-robot system was field tested, which proved its efficacy in inspecting oil and gas assets in shallow waters.

In this paper, a touch less automated face recognition system for smart attendance application was designed using convolutional neural network (CNN). The presented touch less smart attendance system is useful for offices and college’s attendance applications with this the spread of covid-19 type viruses can be restrict. The CNN was trained with dedicated database of 1890 faces with different illumination levels and rotate angles of total 30 targeted classes. A CNN performance analysis was done with 9-layer and 11-layer with different activation functions i.e., Step, Sigmoid, Tanh, softmax, and ReLu. An 11-layer CNN with ReLu activation function offers an accuracy of 96.2% for the designed face database. The system is capable to detect multiple faces from test images using Viola Jones algorithm. Eventually, a web application was designed which helps to monitor the attendance and to generate the report.

This paper reviews the technical aspects of robotic charging for Electric Vehicles (EVs), aiming to identify research trends, methods, and challenges. It implemented the Systematic Literature Review (SLR), starting with the formulation of research question; searching and collecting articles from databases, including Web of Science, Scopus, Dimensions, and Lens; selecting articles; and data extraction. We reviewed the articles published from 2012 to 2022 and found that the number of publications increased exponentially. The top five keywords were electric vehicle, robotic, automatic charging, pose estimation, and computer vision. We continued an in-depth review from the points of view of autonomous docking, charging socket detection-pose estimation, plug insertion, and robot manipulator. No article used a camera, Lidar, or Laser as the sensor that reported successful autonomous docking without position error. Furthermore, we identified two problems when using computer vision for the socket pose estimation and the plug insertion: low robustness against different socket shapes and light conditions; inability to monitor excessive plugging force. Using infrared to locate the socket yielded more robustness. However, it requires modification of the socket on the vehicle. A few articles used a camera and force/torque sensors to control the plug insertion based on different control approaches: model-based control and data-driven machine learning. The challenges were to increase the success rate and shorten the time. Most researchers used commercial 6-DOF robot manipulators, whereas a few designed lower-DOF robot manipulators. Another research challenge was developing a 4-DOF robot manipulator with compliance that ensures a 100% success rate of plug insertion.

Automation and human-robot collaboration are increasing in modern workplaces such as industrial manufacturing. Nowadays, humans rely heavily on advanced robotic devices to perform tasks quickly and accurately. Modern robots with computer vision and artificial intelligence are gaining attention and popularity rapidly. This paper demonstrates how a robot can automatically detect an object’s shape, color, and size using computer vision techniques and act based on information feedback. In this work, a powerful computational model for a robot has been developed that distinguishes an object’s shape, size, and color in real time with high accuracy. Then it can integrate a robotic arm to pick a specific object. A dataset of 6558 images of various monochromatic objects has been developed, containing three colors against a white background and five shapes for the research. The designed system for detection has achieved 99.8% success in an object’s shape detection. Also, the system demonstrated 100% success in the object’s color and size detection with the OpenCV image processing framework. On the other hand, the prototype robotic system based on Raspberry Pi-4B has achieved 80.7% accuracy for geometrical shape detection and 81.07%, and 59.77% accuracy for color recognition and distance measurement, respectively. Moreover, the system guided a robotic arm to pick up the object based on its color and shape with a mean response time of 19 seconds. The idea is to simulate a workplace environment where a worker will ask the robotic systems to perform a task on a specific object. Our robotic system can accurately identify the object’s attributes (e.g., 100%) and is able to perform the task reliably (81%). However, reliability can be improved by using a more powerful computing system, such as the robotic prototype. The article’s contribution is to use a cutting-edge computer vision technique to detect and categorize objects with the help of a small private dataset to shorten the training duration and enable the suggested system to adapt to components that may be needed for creating a new industrial product in a shorter period. The source code and images of the collected dataset can be found at: https://github.com/TituShahoriar/cse499B_Hardware_Proposed_System.

Human’s imagination capability provides recognition of unseen environment which should be improved in robots in order to have better mapping, planning, navigation and exploration capabilities in the fields where the robots are utilized such as military, disasters, and industry. The task of completion of a partial scene via estimating the unobserved parts relied on the known information is called scene extrapolation. It increases performance and satisfies a valid approximation of unseen content even if it is impossible or hard to obtain it due to the issues related with security, environment, etc. In this survey paper, the studies related to learning-based scene extrapolation in robotics are presented and evaluated taking the efficiencies and limitations of the methods into account to provide researchers in this field a general overview on this task and encourage them to improve the current studies for higher success. In addition, the methods which use common datasets and metrics are compared. To the best of our knowledge, there isn’t any survey on this essential topic and we hope this survey will compensate this.