Position Estimation using the Radical Axis Gauss Newton Algorithm: Experimental Analysis

Autores: Luis A. Arellano Cruz, Giselle M. Galván Tejada, Rogelio Lozano

Resumen: 
Generally unmanned aerial vehicles are currently equipped with multiple sensors to help them in their navigation. However, the acquisition of the position at every moment is a hard task to be achieved mainly in crowded scenarios where multiple obstacles make difficult the reception of signal coming from satellites (for global navigation systems) or base stations which perform the functions of anchors in some localization systems. Another issue in this kind of scenarios is the low probability to obtain redundancy systems due to the poor received signal associated to obstructions. For the case where there is only the minimum information required to estimate the position, the named radical axis Gauss Newton (RA-GN) algorithm was previously proposed and successfully evaluated by simulation. Now, in this paper, the RA-GN algorithm is applied to a localization system and evaluated experimentally from a measurement campaign conducted in a semi-forest test field. Moreover, we assess the effect of rotating the tag taken as the representation of the vehicle and found that this rotation introduces changes in the position estimation. All our results in terms of root mean square error are compared with those given by a commercial system. Results reported here show that the RA-GN algorithm is able to improve the accuracy of estimation of the commercial system in real conditions of crowded environments, even under complicated situations where the signal is perturbed by the obstacles and under different angular positions of the tag relative to the anchors.

Revista: Journal of Intelligent & Robotic Systems

https://doi.org/10.1007/s10846-022-01779-x