*Article* **Stratified Particle Filter Monocular SLAM**

**Pawel Slowak \* and Piotr Kaniewski**

> Faculty of Electronics, Military University of Technology, ul. gen. S. Kaliskiego 2, 00-908 Warsaw, Poland; piotr.kaniewski@wat.edu.pl

**\*** Correspondence: pawel.slowak@wat.edu.pl

**Abstract:** This paper presents a solution to the problem of simultaneous localization and mapping (SLAM), developed from a particle filter, utilizing a monocular camera as its main sensor. It implements a novel sample-weighting idea, based on the of sorting of particles into sets and separating those sets with an importance-factor offset. The grouping criteria for samples is the number of landmarks correctly matched by a given particle. This results in the stratification of samples and amplifies weighted differences. The proposed system is designed for a UAV, navigating outdoors, with a downward-pointed camera. To evaluate the proposed method, it is compared with different samples-weighting approaches, using simulated and real-world data. The conducted experiments show that the developed SLAM solution is more accurate and robust than other particle-filter methods, as it allows the employment of a smaller number of particles, lowering the overall computational complexity.

**Keywords:** SLAM; autonomous navigation; particle filter; monocular camera; IMU; UAV
