Sensor Functional Clustering for Data Reducing at Making Robot Control Decisions

This work presents a method for reducing data generated by the sensor system of distributed ground robots by choosing a limited set of embedded devices measuring environmental parameters for the implementation of the current task. Sensor systems of robots with a high degree of accuracy and discretization measure many heterogeneous environmental parameters for making control decisions. Part of the received data is redundant for the functioning of robots, but computing resources are spent on its processing. The analysis of existing approaches for reducing the number of data generated by sensory systems devices made it possible to single out two main approaches. The first one consists of the formation of functional clusters of used devices based on various criteria: sufficiency of energy resources, functional purpose of the devices in the system. The second approach is to pre-aggregate and process data on functional clusters, using data bundling. The proposed method of forming functional clusters and minimizing the amount of data generated by the devices of the sensor system is based on reducing the operating time of the sensor and actuators selected in the cluster to solve the target task. When implementing the method, algorithms were created for making a decision to replace devices in case of failures, algorithms for planning the trajectory of the movement of robots. The results of experimental verification of the method and algorithms in the Gazebo simulator are presented.