This work presents a vision system based on the YOLO algorithm to identify static objects that could be obstacles in the path of a mobile robot. In order to identify the objects and its distances, a Microsoft Kinect sensor was used. In addition, a Nvidia Jetson TX2 GPU was used to increase the image processing algorithm performance. Our experimental results indicate that the YOLO network has detected all the predefined obstacles for which it has been trained with good reliability and the calculus of the distance using the depth information returned by the Microsoft Kinect camera had an error below of 3,64%.

The YOLO and SSD algorithms are tools widely used for detecting objects in images or videos. This is due to the speed of detection and good performance in the identification of objects. This article presents a comparison of the YOLOv3 and SSD MobileNet v2 algorithms for identifying objects in images through simulations, the dataset used is an indoor robotics dataset. In order to reach the objective, several training sessions were carried out to analyze the behavior of each model when detecting objects in images. After analyzing the results, a better performance of the YOLOv3 model was observed, although this model takes more time to complete the training for the same number of steps compared to the SSD MobileNet v2 model. It is worth mentioning that this work presents for the first time a comparison between the SSD MobileNet v2 and YOLOv3 algorithms.

The goal of this work is to analyze and compare trajectory planners for a mobile robot in Robot Operating System (ROS), focusing on the performance of local planners on symmetric and asymmetric environments. In addition, two global planners, Dijkstra and A-star, are implemented in order to have a complete analysis and comprehension of the navigation architecture. Two local planning algorithms, Dynamic Window Approach and Timed Elastic Bands, are analyzed and compared more in depth using the mobile robot TurtleBot 3 Burger, an open-source and low-cost platform. The analyzed criteria were geometric and angular precision of the final position and orientation, time and distance of the complete trajectory, and usage of computational power. Experiments were carried out in two environments with different spatial arrangement of obstacles, with the intention of analyzing the behavior both in simulation with the Gazebo software and in the real robot. Both local planning algorithms enabled the robot to reach the target destination without any collisions, presenting the main difference in the usage of processing power.

In this paper, we use data from the Microsoft Kinect sensor that processes the captured imageof a person using and extracting the joints information on every frame. Then, we propose the creation ofan image derived from all the sequential frames of a gesture the movement, which facilitates training in aconvolutional neural network. We trained a CNN using two strategies: combined training and individualtraining. The strategies were experimented in the convolutional neural network (CNN) using theMSRC-12 dataset, obtaining an accuracy rate of 86.67% in combined training and 90.78% of accuracyrate in the individual training.. Then, the trained neural network was used to classify data obtained fromKinect with a person, obtaining an accuracy rate of 72.08% in combined training and 81.25% inindividualized training. Finally, we use the system to send commands to a mobile robot in order to controlit.