On the comparison of uncertainty criteria for active SLAM

Abstract

ResumenEn este reporte se estudia el cálculo del criterio de optimalidad D, para el caso en que es utilizado como una medida de la incertidumbre de un sistema SLAM. Propiedades del uso de este criterio de medida de la incertidumbre en el contexto de SLAM activo son presentadas, al igual que una comparación contra otros criterios de medida de la incertidumbre tales como la entropía y el criterio de optimalidad A. En este reporte se muestra que contrario a lo divulgado previamente en la literatura científica relacionada, el criterio de optimalidad D es capaz de proporcionar informaciónútil acerca de la incertidumbre que tiene un robot que ejecuta un algoritmo de SLAM. Finalmente, a través de varios experimentos con robots reales y simulados, damos soporte a nuestras afirmaciones y mostramos que el uso del criterio de optimalidad D tiene efecto deseables en varias tareas que hacen uso de algoritmos de SLAM como mapeo y navegación activa. AbstractIn this report, we consider the computation of the D-optimality criterion as a metric for the uncertainty of a SLAM system. Properties regarding the use of this uncertainty criterion in the active SLAM context are highlighted, and comparisons against the A-optimality criterion and entropy are presented. This report shows that contrary to what has been previously reported in the literature, the D-optimality criterion is indeed capable of giving fruitful information as a metric for the uncertainty of a robot performing SLAM. Finally, through various experiments with simulated and real robots, we support our claims and show that the use of D-opt has desirable effects in various SLAM related tasks such as active mapping and exploration.

Fast minimum uncertainty search on a graph map representation

“…In [22] a novel computation form of the uncertainty criterion based on the determinant of the covariance matrix is presented. There the D-opt is computed as follows:…”

A critique of current developments in simultaneous localization and mapping

“…This is known as the active SLAM problem which is addressed in various works. 18,[58][59][60][61][62] …”

Perception-driven navigation: Active visual SLAM for robotic area coverage

“…where the sixth root of the 6-degree of freedom (DOF) pose determinant is used [31], [32] to yield a measure with SI units of m · rad. Essentially, PDN compares the two propagated uncertainties from revisiting and exploring, and then chooses the smaller one whenever the exploration uncertainty exceeds the reference allowable uncertainty.…”