This paper presents the design and implementation of COVID-Bot, an open-source robotic platform for sanitizing single plant environments such as offices, houses, apartments, among others. This development seeks to create a tool that allows contributing to the global fight against the COVID-19 pandemic, from a low-cost and easy-to-replicate robot, which disinfects surfaces through type C ultraviolet radiation. The platform is based on a differential robotic base, an RGB-D camera, a tracking camera, three UV-C lamps, and an embedded computer running the ROS-based control software. The description of the hardware used, the software implemented, and the tests carried out to corroborate the operation of the integrated system are presented. These tests demonstrated that the system is adequate to autonomously cover a one-floor apartment, based on the theoretical radiation distance of the used lamps.

This paper presents the design and development of a cost-oriented mobile robot for domestic assistance. The design of the mechanics, electronics and software necessary for the operation of the robot is detailed, as well as the tests carried out to ensure its correct operation. Concluding that the controllers, kinematics, and odometry work correctly providing a measurement that is close to the real one. The developed robot can be easily equipped with multiple actuators to make it useful in different social tasks, such as: elderly supervision robot, medicine supply robot, companion robot for dependent people, telecare robot, among others. The robot design is focused on allowing its easy replication to be freely used in a large number of domestic applications. As future work, the development of a SLAM algorithm that allows the robot to operate autonomously is proposed.

Digitalization in the fashion industry is attracting the attention of both consumers and online shopping services. Therefore, a personalized and efficient recommendation system is becoming increasingly important. However, most of the traditional systems focus on recommendations without considering the outfit-user relationship, decreasing the accuracy of the recommendations. Therefore, we propose a fashion recommendation system based on user preferences. The adaptive capacity of the system is given by two phases. The first one generates a short-term memory that is constantly updated with the user's interactions. The second one creates a long-term memory based on a DNN. The recommendation system is structured in 3 stages: Database Generator, Model Ranking, and implicit profiling.

The Database Generator encodes the visual characteristics of the garments. The Ranking Model deals with the scoring of the recommendations. The implicit profiling updates the ranking according to the user's preferences. Finally, the system is evaluated using images provided by the user. Through experiments based on user interaction, the system demonstrates adaptation capabilities by recommending similar outfits to the previous user selections. The proposed system demonstrated the ability to adjust to user preferences through human-machine interactions, as required for this type of recommendation system.

We show the application of inverse reinforcement learning (IRL) in discrete and continuous environments based on the apprenticeship focus. The objective is to learn a mathematical definition of a task based on trajectories made by an expert agent. To achieve this, there must be features functions that in some way describe abilities that the agent can learn. Therefore, the description of a task can be formulated as a linear combination of those functions. This learning method was applied in Open-AI gym environments to show the learning process of a task using demonstrations.

This paper presents the theoretical background and the implementation of a Long Short-Term Memory (LSTM) Neural Network architecture to recognize arm movements from video clips. The pose points (corresponding to the position of six body parts: shoulders, elbows and wrists) are extracted with a pre-trained Convolutional Pose Machine. Those points generate sequences over time with 66 (x,y) pairs, which are the input for a neural network, to classify them in 20 movement classes. Our architecture has 128 LSTM cells and presented 92.5% of accuracy on testing data and an execution time of around 6.64ms.

Moreover, we present the methodology used to create our dataset, with 2400 samples of 20 different arms movements, recorded by 6 persons with different physical appearance in a controlled environment.

This paper proposes a system to analyze urban traffic through the using of artificial vision, in order to get reliable information about the traffic flow in cities with severe traffic jam, as in Bogotá, Colombia. It was proposed a method efficient enough to be implemented in an embedded system, in order to process the images captured by a local camera and send the synthesized information to the cloud. This approach would allow spending fewer data transference, because it would not be necessary to send the video of each camera in the city via streaming, instead, each camera would send only the relevant traffic information. The system is able to calculate traffic flow, classified in motorbikes, buses, microbuses, minivans, sedans, SUVs and trucks.

detection was implemented using a cascade classifier that evaluates HAAR features, providing a detection rate of 74.9% and a false positive rate of 1.4%. A Kalman filter was used to track and count the detected vehicles. Finally, a Convolutional Neural Network performing as classifier, with accuracies around 88%. The complete system presented errors around 2.5% in contrast with the manual counting in traffic of Bogotá, Colombia.

We present the application of multiagent reinforcement learning to the problem of traffic light signal control to decrease travel time. We model roads as a collection of agents for each signalized junction. Agents learn to set phases that jointly maximize a reward function that encourages short vehicle queuing delays and queue lengths at all junctions. The first approach that we tested exploits the fact that the reward function can be splitted into contributions per agent. Junctions are modeled as vertices in a coordination graph and the joint action is found with the variable elimination algorithm. The second method exploits the principle of locality to compute the best action for an agent as its best response for a two player game with each member of its neighborhood. We apply the learning methods to a simulated network of 6 intersections, using data from the Transit Department of Bogotá, Colombia. These methods obtained significant reductions in queuing delay with respect to the fixed time control, and in general achieve shorter travel times across the network than some other reinforcement learning based methods found in the literature.

We present a demonstration of two coordination methods for the application of multiagent reinforcement learning to the problem of traffic light signal control to decrease travel time. The first approach that we tested exploits the fact that the reward function can be splitted into contributions per agent. The second method computes the best response for a two player game with each member of its neighborhood. We apply both learning methods through SUMO traffic simulator, using data from the Transit Department of Bogotá, Colombia.

This document shows the results obtained experimentally for the deduction of transfer functions of a tanks plant connected in cascade through measurement of level and temperature variables. To obtain the characteristic curves that describe the behavior of these variables were used sensors installed in the process. Also, the design procedure of a multivariable controller with integral action is described. The digital controller is implemented in real time using the Toolbox Simulink® and software tools from Matlab®, obtaining static and dynamic responses of successful operation to ensure system operation to a temperature and level desirables, applied to production processes.