El trabajo se basó en la adquisición y transmisión de la imagen usando herramientas de simulación como Webots y Choreographe. Este trabajo tuvo como objetivo mejorar la percepción visual del robot humanoide NAO utilizando estos sistemas embebidos que ejecutan una red neuronal convolucional. Como alternativa a unidades de cómputo secuencial se encuentran los Field Programmable Gate Arrays y las Graphics Processing Unit, que tienen un alto grado de paralelismo y bajo consumo de energía.
A pesar de sus capacidades, su sistema de computación secuencial limita la ejecución de algoritmos computacionalmente costosos, como las redes neuronales convolucionales, que han demostrado buen rendimiento en tareas de reconocimiento. Los robots humanoides encuentran aplicación en tareas de interacción humano-robot. Keywords: Convolutional neural networks, field programmable gate array, system-on-a-chip, high-level synthesis, humanoid robot. The results demonstrated that the use of these embedded systems and convolutional neural networks can provide humanoid robots such as NAO with greater visual recognition in tasks that require high accuracy and autonomy. Real-time metrics were obtained when Inception V1, Inception V3 transfer-learning and AlexNet were run on the Ultra96 and Jetson TX2 cards, with frame rates between 28 and 30 frames per second. Xilinx® Ultra96™, Intel® Cyclone® V-SoC and NVIDIA® Jetson™ TX2 cards were used, and Tinier-YOLO, AlexNet, Inception-V1 and Inception V3 transfer-learning networks were executed. In each embedded system, an object recognition stage is performed using commercial convolutional neural network acceleration frameworks.
The methodology adopted here is based on image acquisition and transmission using simulation software: Webots and Choreographe. This study aims to improve the visual perception of a humanoid robot called NAO using these embedded systems running a convolutional neural network. As an alternative to sequential computing units, Field-Programmable Gate Arrays and Graphics Processing Units have a high degree of parallelism and low power consumption. However, despite their capabilities, their sequential computing system limits the execution of computationally expensive algorithms such as convolutional neural networks, which have demonstrated good performance in recognition tasks. Humanoid robots find application in human-robot interaction tasks.
0 kommentar(er)
