Real-time control for robotics is a popular research area in the reinforcement learning community. Through the use of techniques such as reward shaping, researchers have managed to train online agents across a multitude of domains. Despite these advances, solving goal-oriented tasks still requires complex architectural changes or hard constraints to be placed on the problem. In this article, we solve the problem of stacking multiple cubes by combining curriculum learning, reward shaping, and a high number of efficiently parallelized environments. We introduce two curriculum learning settings that allow us to separate the complex task into sequential sub-goals, hence enabling the learning of a problem that may otherwise be too difficult. We focus on discussing the challenges encountered while implementing them in a goal-conditioned environment. Finally, we extend the best configuration identified on a higher complexity environment with differently shaped objects.
One of the major benefits of quantum computing is the potential to resolve complex computational problems faster than can be done by classical methods. There are many prototype-based clustering methods in use today, and the selection of the starting nodes for the center points is often done randomly. Clustering often suffers from accepting a local minima as a valid solution when there are possibly better solutions. We will present the results of a study to leverage the benefits of quantum computing for finding better starting centroids for prototype-based clustering.
Simulation-to-real is the task of training and developing machine learning models and deploying them in real settings with minimal additional training. This approach is becoming increasingly popular in fields such as robotics. However, there is often a gap between the simulated environment and the real world, and machine learning models trained in simulation may not perform as well in the real world. We propose a framework that utilizes a message-passing pipeline to minimize the information gap between simulation and reality. The message-passing pipeline is comprised of three modules: scene understanding, robot planning, and performance validation. First, the scene understanding module aims to match the scene layout between the real environment set-up and its digital twin. Then, the robot planning module solves a robotic task through trial and error in the simulation. Finally, the performance validation module varies the planning results by constantly checking the status difference of the robot and object status between the real set-up and the simulation. In the experiment, we perform a case study that requires a robot to make a cup of coffee. Results show that the robot is able to complete the task under our framework successfully. The robot follows the steps programmed into its system and utilizes its actuators to interact with the coffee machine and other tools required for the task. A noteworthy observation from the experiment is the speed and accuracy with which the robot completed the task. The robot can make a cup of coffee relatively quickly compared with traditional robot planning and control methods, and its movements were precise and efficient. Overall, the results of this case study demonstrate the potential benefits of our method that drive robots for tasks that require precision and efficiency. Further research in this area could lead to the development of even more versatile and adaptable robots, opening up new possibilities for automation in various industries.
In this paper, a position-based visual servo control scheme for a bottle packaging process using the ScorBot-ER-4U robot manipulator is proposed. The control scheme considers the eye-to-hand configuration and is based on the kinematic model of the manipulator robot and the perspective projection model of the vision camera (Pinhole Model). The proposed scheme was evaluated on a virtual environment developed in the Unity3D graphics engine; and experimentally with the ScorBot-ER-4U manipulator robot and the ZED 2 stereo vision camera. Finally, it is concluded that the results obtained by simulation and experimentally show that the control errors converge to zero asymptotically.
Vehicle classification is an essential part of intelligent transportation systems (ITS). This work proposes a model based on transfer learning, combining data augmentation for the recognition and classification of local vehicle classes in Canada. It takes inspiration from contemporary deep learning (DL) achievements for image classification. This makes use of the Dataset named Stanford AI of Vehicles, which has 16185 images. The images in this section are divided into 196 types of vehicles. To increase performance further, additional classification blocks are added to the residual network (ResNet-50)-based model which is being used. In this case, vehicle type details are automatically extracted and classified. A number of measures like accuracy, precision, recall, etc. are used during the analysis to evaluate the results. The proposed model exhibits increased accuracy despite vehicles’ different physical characteristics. In comparison to the current baseline method and the two pre-trained DL systems, AlexNet and VGG-16, our suggested method outperforms them all. The suggested ResNet-50 pre-trained model achieves an accuracy of 90.07
In this paper, we present a modified algorithm based on topological data analysis (TDA) for object localization, and compare its performance with two well-known supervised models, Vision Transformer (ViT) and Yolov7, on two different datasets. Our TDA-based approach returns an IOU (intersection over union) score of 64
Sheaf theory is a potent but intricate tool that is supported by topological theory. It offers more accuracy and adaptability than traditional graph theory when modeling the connections between several characteristics. This is especially valuable in air quality monitoring, where sudden changes in local dust particle density can be hard to measure accurately using commercial instruments. Conventional air quality measurement techniques often depend on calibrating the measurement with standard instruments or calculating the measurement’s moving average over a fixed period. However, this can result in an incorrect index at the measurement location, as well as an excessive smoothing effect on the signal. To address this issue, this study proposes a self-correcting algorithm that employs sheaf theory to account for vehicle counts as a local air quality change-causing factor. By deducing the number of vehicles and incorporating it into the recorded PM2.5 index from low-cost air monitoring sensors, we can achieve real-time self-correction. Additionally, the sheaf-theoretic approach enables straightforward scaling to multiple nodes for further filtering effects. By integrating sheaf theory into air quality monitoring, we can overcome the limitations of conventional techniques and provide more precise and dependable results.
A significant proportion of individuals use freelance platforms as a primary or as a secondary source of their income. However, the majority of freelance platforms have their fair share of problems that need to be addressed. Although, most of these problems have been outlined in recent research papers, but the solutions proposed by these research papers do not take user feedback into consideration. Hence, the purpose of this research paper is to conduct usability evaluation of freelance platforms, that is entirely based on the user's feedback. In this study, an experiment of retrospective testing was conducted with users along with a system usability scale, the results of which were used to determine the necessary design elements of a freelance platform and the main problems that the users faced.
5G and beyond 5G wireless communication technologies need high data rate, low latency, high capacity, and high spectrum efficiency. An adaptive and reconfigurable communication system called cognitive radio (CR) is capable of automatically identifying and utilizing unutilized spectrum while minimizing detrimental interference to the existing system for effective spectrum utilization. Researchers in the field were driven by this advantage of CR. Filter bank multicarrier (FBMC) has been brought into consideration as an alternative to OFDM for 5G systems as a result of this shortcoming. The universal filtered multicarrier (UFMC), however, combines the benefits of FBMC with OFDM due to the large overhead in FBMC. It enhances CR’s spectrum sensing capabilities, making it a good choice for 5G and beyond 5G communication systems. The effective utilization of spectrum and interference management depend on spectrum sensing. The performance of the suggested system is examined, along with the effects of sample size and signal to noise ratio (SNR) values. This study proposed and investigated cooperative spectrum sensing (CSS) using UFMC-based CR for 5G systems. In the cases of AND, OR, and Majority Vote (MV) fusion rules, the effect of adding more CR nodes on system performance is also examined, and improved probability of detection performances are obtained.
This paper presents MF-SET, a multi-task framework for sentiment analysis and aspect-opinion extraction in student evaluations of teaching (SET) in Spanish, which is an understudied area. We first prepared a novel aspect-opinion extraction dataset in Spanish that evaluates nine different aspects of teaching in both positive and negative light. We then developed a multi-task learning framework, including opinion segmentation, multi-class classification, and multi-label classification models for extracting information from Spanish text. The framework uses text generative abilities of GPT-3 combined with BERT to deal with the task. For multi-class classification, we used BERT with different feature sets and sentiments about teachers with respect to their recommendation, intellectual challenges, and learning guidance. Our results show that the model’s performance varies depending on the feature sets and classifiers. Opinion segmentation and multi-label classification were fine-tuned on GPT-3 for the best results of 0.68 Micro F1 for positive and 0.72 for negative aspects.
Smart cities are built on top of Internet of Things (IoT)-based networks, with the aim of improving the quality of life for their citizens and urban surroundings through the provision of smart solutions across numerous applications with streamlined communications. However, due to the complexity of devices involved in forming IoT networks, especially sensors, which play an important role in data acquisition and collecting sensitive information. However, there are various security threats and privacy issues with the smart city infrastructure and the processing of data. Moreover, due to the heterogeneous nature of devices and a lack of suitable security mechanisms and deployment scenarios, traditional security mechanisms are not feasible in the IoT spectrum. Also, the inclusion of malicious nodes and tampered data not only reduces the quality of service but also degrades the network's performance and lifetime. Blockchain, with its decentralized nature and distributed design, can play a significant role in securing IoT-based networks, specifically by providing trustworthy mechanisms for data acquisition. To this end, in this paper, we propose a distributed blockchain-based trustworthy scheme (DBTS) for a network of untrusted Internet of Things (IoT) devices. The suggested approach incorporates smart contracts into heterogeneous Internet of Things (IoT) networks to verify data's veracity.
Teachers and students faced a huge challenge when the first confinements were put in place. COVID-19 created a difficult situation, both economically and socially. The isolation of the children at home had a negative impact on their development. Compared to other pandemics, today we have technology that allows us to teach at a distance. However, not all children and teachers were prepared. They lacked digital skills, equipment, and Internet access. It is in this context that the project HOME 2.0 Education at your fingertips is developed. This way, for two years, there were meetings between teachers who collaborated in the project, sharing knowledge and good practices. In the end, a teacher-to-teacher sharing platform was developed. Its main objective is to share resources to overcome difficulties caused by not having access to school. This paper presents the implementation process of the platform, for which it was decided to perform an Analysis for Requirements with the project members. Through the platform it is possible to submit apps and documents considered relevant by teachers. For those who are looking for these resources, the platform provides a simple way to search and find an app or document that can help in their teaching practice. It was evaluated by teachers from several countries associated with the project.
This study presents a multi-layered microgrid system with an optimization-based energy management system, where the impact of renewable energy penetration and data loss in battery command is investigated. Data loss in battery command can cause voltage instability, energy supply loss, and increased operational costs in microgrid systems, especially in electricity markets. The simulation results show that on average, more data loss results in higher operational costs, but there are situations where less data loss can be more detrimental to microgrid operation than higher levels of data loss. This research provides valuable insights into the effects of data loss in battery command and its potential economic impact on microgrid operation.
This paper presents an agent-based model of malaria transmission taking into account mosquito bites protection and the malaria intra-host treatment. The intra-host treatment is managed by a mathematical model describing the population dynamics of P. falcipharum parasites. The Unified Modeling Language (UML) and the Agent Modeling Language (AML) are used to describe interactions between humans, Anopheles mosquitoes and water ponds. In the initial phase of a 30-day simulation, we placed 500 human agents and 50 water pond agents into the Netlogo development platform. We observed malaria infection cases as a function of human behaviors like treatments, protection with bed nets, spraying with insecticides. The results show that the protection against mosquito bites leads to a reduction in the number of human infection cases for humans and mosquitoes infected cases. The earlier treatment shows the high decrease in the infection rate at the end of simulation. The agent-based model could help public health ministries defining strategies for fighting against malaria transmission.
The transformations that are taking place in the electricity sector are making the electricity grid evolve towards a more intelligent and efficient model known as Smart Grid, in this context, the distributed networks that are integrated into the electrical network are required to control and monitor their total generation. In this project it is proposed, implement a measurement and control system to a photovoltaic plant, which allows the acquisition, transmission, and storage of data, applied to real-time cloud monitoring of a decentralized photovoltaic system. The proposal focuses on a system that articulates the concepts to information system and multi-user remote system using free hardware (Arduino) and Internet of Things (IoT). The system can store the data and communicate with a server in the cloud, the measured variables are the photovoltaic voltage and current. The generation and transmission of electrical energy are not far from the processes of systematization and automation.
This paper presents a control scheme for navigation tasks of an aerial manipulator robot. The proposed controller prioritizes the kinematics of the system considering its high redundancy, which is composed of an aerial platform and an anthropomorphic 3DOF robotic arm, the proposed control scheme is decoupled, i.e., a task is defined for the aerial robot and another task for the robotic arm. To validate the proposed controller, different tests will be performed in a virtual environment and in a partially structured environment. To perform the simulation tests, a virtual environment is developed to visualize the behavior of the manipulator robot, in the simulation environment tasks are planned in the workspace and adjust the controllers, avoiding damage to the physical robot. Once the controllers have been adjusted and simulated, experimental tests are carried out with the aerial manipulator robot.
Detecting adversaries and their intentions during the intelligence gathering step of the attack lifecycle will provide defenders with a strategic advantage. During this step, network scanning tools are a primary resource used by attackers to discover hosts and enumerate services. Tool capabilities and their intent vary and range from scanning for specific services and specific vulnerabilities to large-scale information extraction. By detecting specific tools used during scanning, a defender can infer, to a certain extent, the intentions of an attacker and react accordingly by invoking defenses like dynamic redirection, service blocking, and customized and adaptive honeypots. This paper describes the GEM (Generate, Examine, and Match) system, which implements an automated pipeline mechanism to create rules for intelligence gathering tools. GEM starts by running and collecting data for the tools. It then extracts signatures using differential packet analysis, and finally, it creates Suricata intrusion detection system rules. We tested the system against several scanning tools available on the Kali Linux operating system, totaling 54 configurations. Our analysis shows that the GEM can generate rules for all of the tool configurations. All plaintext configurations can be uniquely identified, and all but six of the 21 encryption configurations can be uniquely identified.
The product of this research is TeaBot, an AI bot that applied GPT-3 models and uses methods of Cognitive Behavioral Therapy (CBT) to help users recognize and challenge distorted thoughts. Considering the limitations of GPT-3 in context understanding, technology was fine-tuned on the self-gathered dataset and all models were tested to find out the most accurate and cost-effective model. Curie model demonstrated the highest performance for the recognition task, and davinci showed the best results in generating a response to users. Additionally, the bot was validated through the 8-week experiment on 68 university students and interviews with practicing psychologists. Research findings revealed that TeaBot is an instrument that can be used for the prevention and intervention of mental disorders with the group assigned to communicate with the bot exhibiting statistically significant differences compared to the control group. The development process also included a manual for the future practice of augmented therapy as well as inform the users who are new to counseling what therapeutic methods were utilized. In conclusion, TeaBot continues to be one of the earliest known bots in Central Asia that applied OpenAI's models and was tested using ethical research methods.
The current era of contactless operations and transactions, the use of Quick Response or QR codes has become more important than before for delivering information to a large group of people. Marketing enterprises commonly use QR codes to quickly deliver information to the intended customers. However, malicious actors also use QR codes to direct potential victims to malicious websites. This research evaluates the potential use of QR codes for delivering malware to targeted users. The objective of the study is to examine if participants who have some cybersecurity awareness training identify fraudulent QR codes. The study was also designed to analyze if the subject of the QR code and poster design affect the ability to determine the validity of QR code. Thirty college students who have taken some level of computer security courses participated in the study. The participants were tasked with examining the poster carefully and scanning the QR code with their phones. They were told to not go to the websites that the QR code contained as this would reveal the validity of the link. Overall, this study shows that participants are very likely to trust QR codes as legitimate source of information even if they are not. This could expose the user to a litany of possible cyber-attacks.
As cloud computing becomes increasingly popular, the need for professionals with cyber security expertise for cloud platforms has become crucial. Continuous education initiatives are vital for ensuring that professionals stay current on the latest advances in cyber security for cloud computing. In this context, synchronous online workshops have emerged as a popular medium for delivering such educational programs, particularly given the flexibility and convenience they offer to learners. This study investigates the effectiveness of the Context Challenge Activity Feedback (CCAF) framework in promoting learners' engagement to maximize learning in a synchronous online workshop on the Fundamentals of Cyber Security for Cloud Computing. It utilizes a mixed-methods approach to analyze post-workshop survey responses, colleagues' feedback, and facilitator observations. Findings indicate that the CCAF framework successfully engages learners, fosters collaboration, and improves learning outcomes. Participants reported satisfaction and a likelihood to recommend the workshop. The study contributes insights into applying the CCAF framework in synchronous online workshops, particularly in cyber security and cloud computing education, and highlights the implications for instructional design and facilitation. Future research should explore the CCAF framework's adaptability to various learning environments, instructional modalities, and the role of technology in engagement and collaboration.
