A chatbot or conversational agent is a software that can interact or ``chat'' with a human user using a natural language, like English, for instance. Since the first chatbot developed, many have been created but most of their problems still persist, like providing the right answer to the user and user acceptance itself. Considering such facts, in this work, we present a chatbot-building framework that considers the use of sentiment analysis and tree timelines to provide a better chatbot answer. For instance, as presented in our experiments, the user can be addressed to a human attendant when its sentiment is very negative, or even try another branch of the tree timeline, as an alternative answer, whenever the user sentiment is less negative.
Multi-objective decision-making in multi-agent scenarios poses multiple challenges. Dealing with multiple objectives and non-stationarity caused by simultaneous learning are only two of them, which have been addressed separately. In this work, reinforcement learning algorithms that tackle both issues together are proposed and applied to a route choice problem, where drivers must select an action in a single-state formulation, while aiming to minimize both their travel time and toll. Hence, we deal with repeated games, now with a multi-objective approach. Advantages, limitations and differences of these algorithms are discussed. Our results show that the proposed algorithms for action selection using reinforcement learning deal with non-stationarity and multiple objectives, while providing alternative solutions to those of centralized methods.
Three-phase induction motors are widely used in different applications in the industry due to their robustness, low cost, and reliability. Untimely identification and correct diagnosis of incipient faults reduce cost and improve the maintenance management of these machines. This paper explores a new method for robust classification of rotor failures in three-phase induction motors (MITs) connected directly to the electrical network, operating in a steady-state, under unbalanced voltages and load conditions. Through an innovative methodology, an analysis of the electrical current signals from 1 hp and 2 hp motors in the frequency domain was performed. Such analysis was applied in constructing input matrices for a Multilayer Perceptron Neural Network (MLPNN) to detect faults. Furthermore, this methodology proved to be robust because the samples of the failing and healthy motors include voltage unbalance conditions in the electrical supply and a significant variation in the load applied to the motor shaft. Such load variation was used for the detection of failures of 1, 2, and 4 broken bars consecutively on the rotor and in the condition of 2 broken bars and 2 other broken bars diametrically opposite. The results were promising and were obtained using 847 real samples from an experimental bench used to construct the neural model and its respective validation.
The World Health Organization (WHO) has declared the novel coronavirus (COVID-19) outbreak a global pandemic in March 2020. Through a lot of cooperation and the effort of scientists, several vaccines have been created. However, there is no guarantee that the virus will shortly disappear, even if a large part of the population is vaccinated. Therefore, non-invasive methods, with low cost and real-time results, are important to detect infected individuals and enable earlier adequate treatment, in addition to preventing the spread of the virus. An alternative is using forced cough sounds and medical information to distinguish a healthy person from those infected with COVID-19 via artificial intelligence. An additional challenge is the unbalancing of these data, as there are more samples of healthy individuals than contaminated ones. We propose here a Deep Neural Network model to classify people as healthy or sick concerning COVID-19. We used here a model composed by an Convolutional Neural Network and two other Neural Networks with two full-connected layers, each one trained with different data from the same individual. To evaluate the performance of the proposed method, we combined two datasets from the literature: COUGHVID and Coswara. That dataset contains clinical information regarding previous respiratory conditions, symptoms (fever or muscle pain), and a cough record. The results show that our model is simpler (with fewer parameters) than those from the literature and generalizes better the prediction of infected individuals. The proposal presents an average Area Under the ROC Curve (AUC) equal to 0.885 with a confidence interval (0.881 - 0.888), while the literature reports 0.771 with (0.752 - 0.783).
This paper presents the performance evaluation of a wireless sensor mesh network, for monitoring temperature and humidity in horticultural products when transported in truck galleys. For this purpose a software solution was proposed using ESP8266 devices powered by batteries. The mesh network was managed by the painlessMesh library. The proposed solution aims to minimize the energy consumption of the sensor nodes. The validation of the solution was performed in an area simulating a galley of a truck, where five sensor nodes and a root node were distributed. The tests were developed considering four different models involving variations in messages delivery confirmation, number of attempts until successful delivery and duty cycle duration of the nodes. The performance evaluation of the solution aimed to determine, connectivity rate, sending rate after connection and delivery rates of the first and second attempts. The results obtained show that the message delivery confirmation does not bring added value to the solution, contributing only to increase energy consumption. The use of synchronous duty cycles also showed worse results than the asynchronous use. These results allow the creation of a knowledge base for the use of this solution in a real context.
This paper presents the proposal, implementation and validation of a low cost fault-tolerant functional prototype for livestock monitoring. This prototype uses IoT devices, ESP8266 and ESP32, creating a mesh network, managed by the painlessMesh library, with WiFi and LoRa technologies. It allows, for instance, the collection of vital signs from animals. In comparison with the traditional method of livestock examination, this cost-efficient approach reduces manual labor and saves working time. It also improves animal health, increases profits and decreases the environmental footprint.
The smart city systems development connected to the Internet of Things (IoT) has been the goal of several works in the multi-agent system field. Nevertheless, just a few projects demonstrate how to deploy and make the connection among the employed systems. This paper proposes an approach towards the integration of a MAS through the JaCaMo framework plus an Urban Simulation Tool (SUMO), IoT applications (Node-RED, InfluxDB, and Grafana), and an IoT platform (Konker). The integration presented in this paper applies in a Smart Parking scenario with real features, where is shown the integration and the connection through all layers, from agent level to artifacts, including real environment and simulation, as well as IoT applications. In future works, we intend to establish a methodology that shows how to properly integrate these different applications regardless of the scenario and the used tools.
We present in this paper a novel approach for measuring Bourdieusian Social Capital (BSC) within Institutional Pages and Profiles. We analyse Facebook's Institutional Pages and Twitter's Institutional Profiles. Supported by Pierre Bourdie's theory, we search for directions to identify and capture data related to sociability practices, i. e. actions performed such as Like, Comment and Share. The system of symbolic exchanges and mutual recognition treated by Pierre Bourdieu is represented and extracted automatically from these data in the form of generalized sequential patterns. In this format, the social interactions captured from each page are represented as sequences of actions. Next, we also use such data to measure the frequency of occurrence of each sequence. From such frequencies, we compute the effective mobilization capacity. Finally, the volume of BSC is computed based on the capacity of effective mobilization, the number of social interactions captured and the number of followers on each page. The results are aligned with Bourdieu's theory. The approach can be generalized to institutional pages or profiles in Online Social Networks.
Recently, with the development of the robot industry, various types of robots have been developed, and robots are being used in various fields as well as industrial fields. Many studies show the possibility of extending the development direction of such companion robots to education and show many research cases. Especially for children, social interaction training during infancy and childhood has a great impact as an adult. Among them, social initiation, which means trying social interaction first, requires sufficient training and interaction experience. So, it is necessary to understand the meaning of social initiation in which the robot attempts social interaction in a situation where the user is concentrating on the task. Learning from robots is effective for infants and young children, and the intimacy formed by social initiation of robots can maximize the learning effect. For those reason, in a 1:1 interaction between the user and the robot, we understand how social initiation of the robot can attract the user's attention when the user is concentrating on the task, and what form of social initiation forms the social intimacy between the user and the robot. In particular, this study conducted an experiment focusing on the change in user's interest and liking according to the language type of the robot among the social initiation types, and the results were derived. This study examined the characteristics of robot social initiation targeting adults first and further work is planned to extend the result to children.
We expect legged robots to perform complex navigation tasks by having higher mobility and sensing capability. However, the mobility does not easily lead to the ability of performing complex user tasks due to restricted interfaces such as a joystick. To perform tasks that users want, the robots require an expressive and accessible interface that can deliver human intention with lower mental demands, even in complex environments. In this work, we propose a novel natural language-guided robotic navigation framework that can effectively ground natural-language commands in large space. Our framework consists of three modules: a scene-graph generator, a grounding network, and a semantic navigation system. The scene-graph generator incrementally stores the semantic information of object instances, properties, and relationships. Then, the proposed scene graph-based grounding network (SGGNet) predicts the desired goal robustly by associating instances in a scene graph with a user command. Finally, the navigation system enables the robot to reach the goal location. Our evaluation result shows SGGNet achieves a grounding accuracy of 77.8% given 3, 000 scene graphs and 9, 000 natural language commands. The model also achieves a grounding accuracy of $$52.4\%$$ given unforeseen objects. We demonstrate the robust performance of the proposed framework in three real-world scenarios with various speech commands.
This paper presents Planar Fitting Transformation (PFT), a highly efficiency-oriented 3D point cloud registration. Based on Normal Distribution Transformation (NDT), we replace the gaussian approximation with least-squared planar fitting, dramatically increasing the FPS to 500 Hz with CPU (18 threads). As an alternative to the time-consuming neighbor search, we propose octomerge voxelization to enhance robustness without compromising efficiency. As a result, our work significantly outperforms all state-of-the-art methods in execution time, while retaining a comparable level of accuracy.
Today, cloud environments are widely used as execution platforms for most applications. In these environments, virtualized applications often share computing resources. Although this increases hardware utilization, resources competition can cause performance degradation, and knowing which applications can run on the same host without causing too much interference is key to a better scheduling and performance. Therefore, it is important to predict the resource consumption profile of applications in their subsequent iterations. This work evaluates the use of machine learning techniques to predict the increase or decrease in computational resources consumption. The prediction models are evaluated through experiments using real and benchmark applications. Finally, we conclude that some models offer significantly better performance when compared to the current trend of resource usage. These models averaged up to 94% on the F1 metric for this task.
Autism Spectrum Disorder (ASD) is a common but complex disorder to diagnose since there are no imaging or blood tests that can detect ASD. Several techniques can be used, such as diagnostic scales that contain specific questionnaires formulated by specialists that serve as a guide in the diagnostic process. In this paper, Machine Learning (ML) was applied on three public databases containing AQ-10 test results for adults, adolescents, and children; as well as other characteristics that could influence the diagnosis of ASD. Experiments were carried out on the databases to list which attributes would be truly relevant for the diagnosis of ASD using ML, which could be of great value for medical students or residents, and for physicians who are not specialists in ASD. The experiments have shown that it is possible to reduce the number of attributes to only 5 while maintaining an Accuracy above 0.9. In the other Database to maintain the same level of Accuracy, the fewer attribute numbers were 7. The Support Vector Machine stood out from the others algorithms used in this paper, obtaining superior results in all scenarios.
The development of an optimal controller for stabilization of a quadrotor system using an adaptive critic structure based on policy iteration schemes is proposed in this paper. This approach is inserted in the context of Approximate Dynamic Programming and it is used to solve optimal decision problems on-line, without requiring complete knowledge of the system dynamics model to be controlled. The main feature of the adaptive critic design method that allows for on-line implementation is that it solves the Bellman optimality equation in a forward-in-time fashion, whereas traditional dynamic programming requires a backward-in-time procedure. This feedback control design technique is able to tune the controller parameters on-line in the presence of variations in plant dynamics and external disturbances using data measured along the system trajectories. Computational simulation results based on a quadrotor model demonstrate the effectiveness of the proposed control scheme.
Beekeeping is one of the most important activities for humans. Since ancient times, honey has been used in the treatment of several diseases and is an extremely powerful antioxidant. The process of visual analysis of the apiary requires trained specialists who try to obtain relevant information to make a decision about what to do with the honeycomb. Since the process is performed manually, given the complexity of the task, opportunities arise for the application of automated systems that can assist the beekeeper's decision making. Thus, this paper presents the development of the application \textit{SegBee}, a computational tool that performs the segmentation in the apiary plates, where there is the presence of honey, in an accessible, fast and practical way. To do this, the OpenCV library was used for the digital image processing part, and the Kivy library was used to develop the interface of the mobile application. The tests performed showed that the images were adequately segmented by \textit{SegBee}, indicating where the honey is located on each analyzed plate. A visual comparison was made between results obtained by \textit{SegBee} and another commercial application, demonstrating the effectiveness of the developed tool. The proposed solution contributes to the improvement of the beekeeping professionals' work, once the application is simple to use and fast to process, being able to help in the honey identification task in apiaries plates.
This article comparatively analyzes two platforms based on Blockchain, aiming at the management of electronic health records in the public healthcare system of Brazil. The difference between the platforms primarily lies in the deployed consensus algorithm. Efficiency, availability, integrity, and confidentiality requirements are evaluated through analytical models and theoretical discussions. Among the obtained results, we highlight the following: (i) the platform with a voting-based consensus algorithm yields a more efficient system, but is more prone to service unavailability, than that of the platform deploying an intensive-compute consensus algorithm; (ii) integrity and confidentiality requirements may be satisfactorily met regardless of the consensus type. As the main contribution, this article provides valuable experimental results and theoretical subsidies, which together complement previous research and help to lay the groundwork for the fruitful development of real projects. Finally, conclusions and future work conclude this article.
Drilling technology is commonly used in resource exploration and foundation construction fields. In particular, new drilling technologies are being researched and developed for efficient exploration for finding new energy resources. However, it has not been able to deviate from the existing framework of the huge equipment size and complex process. In addition, it causes environmental pollution in the process of removing soil debris generated during excavation. Most of the resources emerging as new energy sources are distributed in shallow and wide areas; therefore, technology to efficiently explore them is required. In this paper, a biomimetic embedded drilling robot that mimics the physical structure and digging habits of moles is developed to overcome the limitations of the existing drilling systems. An expandable drill bit that mimics the excavation tool of an African mole-rat and an eco-friendly debris removal system that mimics the forelimbs of a European mole has been developed. The details of the waist and locomotion mechanisms are described in this paper. Through these mechanisms, it is possible to provide directionality and efficient moving under the ground. The feasibility of the mechanisms is evaluated through dynamic simulations, and the performance of the mole robot (Mole-bot) is verified through actual experiments.
Path planning is an essential part of planning in robotics systems. It generates robot trajectory that avoids collision with obstacles. Besides static obstacles, we need to also consider moving obstacles in a dynamic environment. In multi-agent systems, these obstacles often have known trajectory that must be avoided. Related works in sampling-based algorithms proposed a re-planning mechanism to consider the moving obstacle. However, those algorithms assume that the obstacles are fixed in each re-planning step. The paper proposed a novel algorithm based on RRT* that considers moving obstacles that consider known spatial-temporal trajectory of moving obstacles. The simulation result shows that the proposed algorithm can generate collision-free trajectory with comparable performance with baseline RRT*.
Research in the area of Human-robot Interaction (HRI) has gained momentum in recent years. The robot-based intervention system has spread to help the less fortunate specifically children who suffer from Autism Spectrum Disorder (ASD). These robot-based intervention studies utilized HRI platform in improving the impaired skills, such as, social skills, motor skills, and behavior. Recently, robot-based therapies have shown encouraging outcomes in improving the social skills of Autism Spectrum Disorder Children (ASDC). Herein, this paper elaborates a pilot study on initial responses of ASDC when being introduced to a humanoid QTrobot. QTrobot is chosen because of its ability to show facial expression. The ability to subtly show emotion is very important as it is going to be used for our differential reinforcement technique. The pilot experiment consists of 5 simple modules to prompt the responses of the participants. From the results, it clearly can be seen that humanoid QTrobot has a potential to be an HRI platform to initiate interaction among ASDC.
For conventional autonomous vehicle development, image detection plays an essential role in perception tasks. The more comprehensive detection scenarios to cover, the more computing power to consume. In this paper, we proposed an implementation work to demonstrate the traffic signal state broadcasting for an end-to-end autonomous shuttle service. According to the experimental result, the proposed system can significantly offload the image detection task in obtaining the traffic signal state. Additionally, the non-line-of-sight traffic signal state perception is effectively available via the C-V2X wireless communication technique to secure the safety of an end-to-end autonomous shuttle transportation service.
