Deep Learning Models to Reduce Stray Light in TJ-II Thomson Scattering Diagnostic.

Nuclear fusion is a potential source of energy that could supply the growing needs of the world population for millions of years. Several experimental thermonuclear fusion devices try to understand and control the nuclear fusion process. A very interesting diagnostic called Thomson scattering (TS) is performed in the Spanish fusion device TJ-II. This diagnostic takes images to measure the temperature and density profiles of the plasma, which is heated to very high temperatures to produce fusion plasma. Each image captures spectra of laser light scattered by the plasma under different conditions. Unfortunately, some images are corrupted by noise called stray light that affects the measurement of the profiles. In this work, we propose the use of deep learning models to reduce the stray light that appears in the diagnostic. The proposed approach utilizes a Pix2Pix neural network, which is an image-to-image translation based on a generative adversarial network (GAN). This network learns to translateimages affected by stray light to images without stray light. This allows for the effective removal of the noise that affects the measurements of the TS diagnostic, avoiding the need for manual image processing adjustments. The proposed method shows a better performance, reducing the noise up to 98% inimages, which surpassesprevious works that obtained 85% for the validation dataset.

Development and Control of a Real Spherical Robot.

This paper presents the design and implementation of a spherical robot with an internal mechanism based on a pendulum. The design is based on significant improvements made, including an electronics upgrade, to a previous robot prototype developed in our laboratory. Such modifications do not significantly impact its corresponding simulation model previously developed in CoppeliaSim, so it can be used with minor modifications. The robot is incorporated into a real test platform designed and built for this purpose. As part of the incorporation of the robot into the platform, software codes are made to detect its position and orientation, using the system SwisTrack, to control its position and speed. This implementation allows successful testing of control algorithms previously developed by the authors for other robots such as Villela, the Integral Proportional Controller, and Reinforcement Learning.

BERT for Activity Recognition Using Sequences of Skeleton Features and Data Augmentation with GAN.

Recently, the scientific community has placed great emphasis on the recognition of human activity, especially in the area of health and care for the elderly. There are already practical applications of activity recognition and unusual conditions that use body sensors such as wrist-worn devices or neck pendants. These relatively simple devices may be prone to errors, might be uncomfortable to wear, might be forgotten or not worn, and are unable to detect more subtle conditions such as incorrect postures. Therefore, other proposed methods are based on the use of images and videos to carry out human activity recognition, even in open spaces and with multiple people. However, the resulting increase in the size and complexity involved when using image data requires the use of the most recent advanced machine learning and deep learning techniques. This paper presents an approach based on deep learning with attention to the recognition of activities from multiple frames. Feature extraction is performed by estimating the pose of the human skeleton, and classification is performed using a neural network based on Bidirectional Encoder Representation of Transformers (BERT). This algorithm was trained with the UP-Fall public dataset, generating more balanced artificial data with a Generative Adversarial Neural network (GAN), and evaluated with real data, outperforming the results of other activity recognition methods using the same dataset.

Modeling and Control of a Spherical Robot in the CoppeliaSim Simulator.

This article presents the development of a model of a spherical robot that rolls to move and has a single point of support with the surface. The model was developed in the CoppeliaSim simulator, which is a versatile tool for implementing this kind of experience. The model was tested under several scenarios and control goals (i.e., position control, path-following and formation control) with control strategies such as reinforcement learning, and Villela and IPC algorithms. The results of these approaches were compared using performance indexes to analyze the performance of the model under different scenarios. The model and examples with different control scenarios are available online.

Human Activity Recognition by Sequences of Skeleton Features.

In recent years, much effort has been devoted to the development of applications capable of detecting different types of human activity. In this field, fall detection is particularly relevant, especially for the elderly. On the one hand, some applications use wearable sensors that are integrated into cell phones, necklaces or smart bracelets to detect sudden movements of the person wearing the device. The main drawback of these types of systems is that these devices must be placed on a person's body. This is a major drawback because they can be uncomfortable, in addition to the fact that these systems cannot be implemented in open spaces and with unfamiliar people. In contrast, other approaches perform activity recognition from video camera images, which have many advantages over the previous ones since the user is not required to wear the sensors. As a result, these applications can be implemented in open spaces and with unknown people. This paper presents a vision-based algorithm for activity recognition. The main contribution of this work is to use human skeleton pose estimation as a feature extraction method for activity detection in video camera images. The use of this method allows the detection of multiple people's activities in the same scene. The algorithm is also capable of classifying multi-frame activities, precisely for those that need more than one frame to be detected. The method is evaluated with the public UP-FALL dataset and compared to similar algorithms using the same dataset.

Deep learning exoplanets detection by combining real and synthetic data.

Scientists and astronomers have attached great importance to the task of discovering new exoplanets, even more so if they are in the habitable zone. To date, more than 4300 exoplanets have been confirmed by NASA, using various discovery techniques, including planetary transits, in addition to the use of various databases provided by space and ground-based telescopes. This article proposes the development of a deep learning system for detecting planetary transits in Kepler Telescope light curves. The approach is based on related work from the literature and enhanced to validation with real light curves. A CNN classification model is trained from a mixture of real and synthetic data. The model is then validated only with unknown real data. The best ratio of synthetic data is determined by the performance of an optimisation technique and a sensitivity analysis. The precision, accuracy and true positive rate of the best model obtained are determined and compared with other similar works. The results demonstrate that the use of synthetic data on the training stage can improve the transit detection performance on real light curves.

A Distributed Vision-Based Navigation System for Khepera IV Mobile Robots.

This work presents the development and implementation of a distributed navigation system based on object recognition algorithms. The main goal is to introduce advanced algorithms for image processing and artificial intelligence techniques for teaching control of mobile robots. The autonomous system consists of a wheeled mobile robot with an integrated color camera. The robot navigates through a laboratory scenario where the track and several traffic signals must be detected and recognized by using the images acquired with its on-board camera. The images are sent to a computer server that performs a computer vision algorithm to recognize the objects. The computer calculates the corresponding speeds of the robot according to the object detected. The speeds are sent back to the robot, which acts to carry out the corresponding manoeuvre. Three different algorithms have been tested in simulation and a practical mobile robot laboratory. The results show an average of 84% success rate for object recognition in experiments with the real mobile robot platform.

A Neural Network Approach for Building An Obstacle Detection Model by Fusion of Proximity Sensors Data.

Proximity sensors are broadly used in mobile robots for obstacle detection. The traditional calibration process of this kind of sensor could be a time-consuming task because it is usually done by identification in a manual and repetitive way. The resulting obstacles detection models are usually nonlinear functions that can be different for each proximity sensor attached to the robot. In addition, the model is highly dependent on the type of sensor (e.g., ultrasonic or infrared), on changes in light intensity, and on the properties of the obstacle such as shape, colour, and surface texture, among others. That is why in some situations it could be useful to gather all the measurements provided by different kinds of sensor in order to build a unique model that estimates the distances to the obstacles around the robot. This paper presents a novel approach to get an obstacles detection model based on the fusion of sensors data and automatic calibration by using artificial neural networks.

Event-Based Control Strategy for Mobile Robots in Wireless Environments.

In this paper, a new event-based control strategy for mobile robots is presented. It has been designed to work in wireless environments where a centralized controller has to interchange information with the robots over an RF (radio frequency) interface. The event-based architectures have been developed for differential wheeled robots, although they can be applied to other kinds of robots in a simple way. The solution has been checked over classical navigation algorithms, like wall following and obstacle avoidance, using scenarios with a unique or multiple robots. A comparison between the proposed architectures and the classical discrete-time strategy is also carried out. The experimental results shows that the proposed solution has a higher efficiency in communication resource usage than the classical discrete-time strategy with the same accuracy.

A mobile robots experimental environment with event-based wireless communication.

An experimental platform to communicate between a set of mobile robots through a wireless network has been developed. The mobile robots get their position through a camera which performs as sensor. The video images are processed in a PC and a Waspmote card sends the corresponding position to each robot using the ZigBee standard. A distributed control algorithm based on event-triggered communications has been designed and implemented to bring the robots into the desired formation. Each robot communicates to its neighbors only at event times. Furthermore, a simulation tool has been developed to design and perform experiments with the system. An example of usage is presented.