State Estimation for Swarm UAVs Under Data Dropout Condition

Abstract

In this work, a method based on position predicting, velocity filtering and self adaptive parameter tunning is addressed for state estimation and control for swarm of mini unmanned aerial vehicles (UAVs), in order to deal with random noise and data dropout appeared during flights. Under conditions of random data dropout rates and communication latencies, the presented algorithm gives position prediction based on filtered velocity estimation and it fuses the prediction with sensor data. At the same time it corrects the prediction by the error between prediction and measurement of the previous step. The algorithm is designed for tracking mini UAVs with identical marker configuration, and the principles refered is in potential of serving to state estimation in various circumstances. Based on this localization algorithm, a cascade nonlinear control model is developed for swarm UAV control. This work contributes mainly to the object localization and control in a multi-agent system in which all the agents are considered to be in an identical form, hoping that this work will be the testbed for more complicated swarm robot control experiments. Comparison results of state estimation are presented by implementing experiments with or without data dropout.