Pre-processing of observation data of intelligent agents using real-time causal filters.

Intelligent agents in robotics interact with the environment and perceive this environment using sensors. However, the results of any measurements of physical quantities are distorted by random noise, and there is a need to extract information from noisy measurements. For this, existing approaches involve the construction of models for processing existing data. We compared various types of causal filters operating in real time or with a fixed delay: a simple moving average, an exponential moving average with an adaptive coefficient, a simple moving median and a Kalman filter. Also, for the preliminary processing of the observation data of intelligent agents, we investigated a two-stage architecture of a complex filter consisting of a Kalman filter and a simple moving median. A simulation was carried out in which the signal-to-noise ratio varied from 20 dB to 40 dB for N=200,000 implementations of Gaussian noise sequences. The estimation of the root-mean-square error for all the considered filter variants under conditions of high noise of the observed data was carried out, the evaluation of the further application of the processed observation data in the computing wireless network of intelligent agents was carried out. [ABSTRACT FROM AUTHOR]