Avaliação de soluções de inteligência artificial preditiva para um cenário de internet tátil.

Abstract

The internet is evolving to allow physical interactions with real or virtual objects remotely. This new paradigm is known as Tactile Internet, which requires an end-to-end latency of 1 ms. In order to decrease the perception of delay in a tactile internet system, predictive artificial intelligence approaches should be investigated. This work proposes the development of a system for simulation and evaluation of predictive and control artificial intelligence approaches in a Tactile Internet scenario. The basic components of the system are presented, and details regarding the control flow and design of the applications are provided. In addition, modeling and remote control of the manipulator adopted as a slave robot is performed. Two control approaches are proposed for the developed system: Proportional-Derivative (PD) control and Model Predictive Control (MPC). Results of experimental simulations of the control approaches (in the developed system) are presented, validating and emphasizing the performance of the system. The PD controller showed better performance, given the tuning of its parameters by genetic algorithm. Prediction approaches are investigated and a prediction approach based on the Autoregressive Exogenous (ARX) artificial neural network is evaluated using the developed system. Results obtained from the simulation prove the effectiveness of the approach in predicting the response of the slave model on communication delay in the network domain.