Monte Carlo-Based Indoor UAV Localization in Dynamic Environments Using Only Four Onboard Range Sensors

Mohammad Reza Ranjbar Divkoti; A. Pedro Aguiar

cs.ROOct 23, 2026

Monte Carlo-Based Indoor UAV Localization in Dynamic Environments Using Only Four Onboard Range Sensors

Authors: Mohammad Reza Ranjbar Divkoti, A. Pedro Aguiar

Organizations: SYSTEC-ARISE Research Center for Systems and Technologies, Faculty of Engineering, University of Porto 4200-465 Porto, Portugal.

Abstract

Indoor UAV localization with minimal onboard sensing remains challenging, particularly in dynamic environments. We propose augmented filtered Monte Carlo localization (AF-MCL) for 2D localization using only four single-point range sensors and measured heading. AF-MCL rejects range measurements inconsistent with the known static map and adaptively injects particles to recover from localization divergence and kidnapping. Beam classification is performed only at the previous position estimate, making its ray-casting cost independent of the particle count. The particle distribution additionally provides an online measure of localization uncertainty. Real-world experiments in a 2.10x1.63 m environment yield mean position errors of 6.4-7.7 cm under static and dynamic conditions. Simulations in 6x6 m and 10x10 m environments yield 4.5-5.2 cm mean errors; under dynamic obstacles, standard MCL reaches 12.3 cm and reliable MCL 7.7 cm. The results show that robust indoor localization under dynamic obstacles is achievable with only four single-point range measurements and heading.

Figures & tables

Fig. 1 : UAV localization with four directional range sensors in a known static map and unknown dynamic obstacles.

$N$	$z_{\max}$	$z_{\mathrm{r}}$	$z_{\mathrm{short}}$	$z_{\mathrm{hit}}$	$\Sigma_{0}$	$\gamma$	$\lambda_{\mathrm{short}}$	$\Sigma_{u}$	$\sigma_{\mathrm{obs}}$	$\alpha_{\text{slow}}$	$\alpha_{\text{fast}}$	$\chi$	$\nu$
100	0.1	0.1	0.1	0.7	0.2	0.5	4.0	0.03	0.15	0.001	0.9	0.4	0.3

TABLE I : Simulation parameters.

Fig. 2 : Webots simulation environment showing the Crazyflie UAV, static walls, and the dynamic obstacles used to test localization robustness.

Panel of Figure 4 — Fig. 3 : Ablation results under dynamic obstacles: (a–d) estimated vs. ground-truth paths for square and S-shaped trajectories in $6\times 6$ m (a, b) and $10\times 10$ m (c, d) environments; (e–h) corresponding position errors over time.

Env.	Traj.	Method	Mean Pos.	Mean	Mean $\sigma$
[m]			Error [m]	Runtime [ms]	[m]
$6\times 6$	S-shape	MCL	0.075 $\pm$ 0.041	0.68 $\pm$ 0.67	0.048
		A-MCL	0.220 $\pm$ 0.187	1.60 $\pm$ 1.05	0.565
		F-MCL	0.048 $\pm$ 0.032	2.73 $\pm$ 1.35	0.049
		R-MCL	0.052 $\pm$ 0.018	2.10 $\pm$ 1.15	0.049
		AF-MCL	0.047 $\pm$ 0.026	3.15 $\pm$ 1.26	0.051

TABLE II : Position Error and Runtime Performance across Environments, Trajectories, and Methods

Panel of Figure 6 — Fig. 4 : Kidnapping scenario: (a) Trajectories and kidnapping window. (b) Position error over time. (c) $p_{\mathrm{random}}$ particle injection. (d) $\bar{\ell}_{\mathrm{slow}}$ and $\bar{\ell}_{\mathrm{fast}}$ triggering recovery.

$N$	Cls	Prp	Wgt	Rsm	Tot	Err
25	1.79	0.08	0.78	0.18	2.95 $\pm$ 0.92	0.049 $\pm$ 0.036
50	2.04	0.09	0.92	0.29	3.48 $\pm$ 1.10	0.050 $\pm$ 0.045
100	1.95	0.09	0.90	0.46	3.53 $\pm$ 0.81	0.048 $\pm$ 0.023
200	2.09	0.11	1.07	0.83	4.25 $\pm$ 1.16	0.046 $\pm$ 0.028
400	2.02	0.13	1.08	1.51	4.88 $\pm$ 0.76	0.047 $\pm$ 0.025

TABLE III : Runtime and position error vs.

N

. Mean stage times in ms: Cls = beam classification (incl. ray casting), Prp = propagation, Wgt = weighting, Rsm = resampling, Tot = total update time (mean

\pm

std, ms); Err = position error (m);

\sigma

= online uncertainty (m).

Fig. 5 : $\sigma_{\min}(H_{Q_{t}})$ (blue) and position error (red) along the S-shaped trajectory with dynamic obstacles. Drops in $\sigma_{\min}$ from beam rejection coincide with elevated localization error.

Panel of Figure 9 — Fig. 6 : (a) Crazyflie 2.1 with Flow Deck v2 and Multi-ranger Deck. (b) Crazyradio PA used for host-PC communication. (c) Eight Qualisys cameras supplied ground-truth position.

Panel of Figure 10 — Fig. 7 : Real-world test environments: (a) S-shape trajectory with static obstacles, (b) S-shape trajectory with static and dynamic obstacles, (c) square trajectory with static obstacles, and (d) square trajectory with static and dynamic obstacles.

Traj.	Scenario	Mean Pos.	Mean	Mean $\sigma$
		Error [m]	Runtime [ms]	[m]
S-shape	No Obstacle	0.067 $\pm$ 0.012	1.43 $\pm$ 0.24	0.058
	Static Obstacles	0.070 $\pm$ 0.009	1.21 $\pm$ 0.02	0.061
	Static&Dynamic Obstacles	0.064 $\pm$ 0.004	1.21 $\pm$ 0.01	0.068
Square	No Obstacle	0.068 $\pm$ 0.009	1.26 $\pm$ 0.03	0.050
	Static Obstacles	0.073 $\pm$ 0.006	1.21 $\pm$ 0.06	0.051

TABLE IV : AF-MCL localization performance

Panel of Figure 12 — Fig. 8 : Real-world scenario results. (a)–(d) AF-MCL position estimate versus ground truth for: (a) S-shape trajectory with static obstacles; (b) S-shape trajectory with static and dynamic obstacles; (c) square trajectory with static obstacles; (d) square trajectory with static and dynamic obstacles. (e)–(h) Corresponding position error over time for (a)–(d), with mean errors of 0.061 m, 0.059 m, 0.065 m, and 0.076 m, respectively.

Robot Localization

Robot localization estimates a robot’s position and orientation from onboard sensors, maps, and observations of its surroundings.

See all 177 papers

Towards Quadruped-Provided Localization and Active Tracking for Micro-UAVs
Oct 5, 2026Alejandro Lorite Mora, Andrés FaíñaUAV Navigation Cooperative Localization
IT University of Copenhagen, Rued Langgaards Vej 7, Copenhagen 2300, Denmark · Helix Lab, Campus Kalundborg 3, 4400 Kalundborg, Denmark · Novo Nordisk A/S, Hallas Alle 1, 4400 Kalundborg, Denmark
Cross-Session 3D LiDAR and Camera Fusion for Robust Localization of Unmanned Aerial Vehicles in GPS-Denied Environments
Jun 27, 2026Cong Hoang Quach, Chi Thanh Vo, Dong LT. Tran +3UAV Navigation Camera-LiDAR Fusion
1 – School of Electrical and Data Engineering, University of Technology Sydney, New South Wales, Australia · 2 – Institute of Aerospace Engineering and Technology, Duy Tan University, Da Nang, Vietnam · 5 – Department of Artificial Intelligence and Robotics, Sejong University, Seoul, South Korea +2
Learning-Based Dynamic Obstacle Avoidance for a UAV Using Only Three Range Sensors
Sep 13, 2026Mohammad Reza Ranjbar Divkoti, A. Pedro AguiarRobot Navigation UAV Navigation
SYSTEC-ARISE Research Center for Systems and Technologies, Faculty of Engineering, University of Porto, 4200-465, Porto, Portugal.

Oct 5, 2026J/K move · Enter open · S save

Towards Quadruped-Provided Localization and Active Tracking for Micro-UAVs

Alejandro Lorite Mora, Andrés Faíña

IT University of Copenhagen, Rued Langgaards Vej 7, Copenhagen 2300, Denmark · Helix Lab, Campus Kalundborg 3, 4400 Kalundborg, Denmark · Novo Nordisk A/S, Hallas Alle 1, 4400 Kalundborg, Denmark

UAV Navigation Cooperative Localization Robot Localization

Open paper AskPDF arXiv

Micro unmanned aerial vehicles (micro-UAVs) are small enough to reach confined spaces that larger robots cannot access, but too small to carry the sensing and computing power required for autonomous flight. We move the localization stack entirely off the aerial platform onto a quadruped robot with a 7-degree-of-freedom (DOF) arm, which supplies the micro-UAV (27 g bare, 42 g with fiducial markers) its full 6-DOF pose. A camera at the arm's end-effector detects AprilTag fiducial markers on the drone and composes that observation with the quadruped's own self-localization to place the drone in a shared map frame, so the ground robot localizes its partner, rather than only tracking it relative to the camera. The arm acts as an actively-controlled observer, repositioning to keep the drone in view as both robots move; the drone carries only an inertial measurement unit and fuses the external pose to fly commanded setpoints. In lab flights the external pose is accurate to 12-16 mm, enough to fly the drone autonomously within 2-5 cm of motion-capture-fed control. Having the quadruped actively follow the drone reduces the tracking error from 11.0 cm to 6.9 cm by holding the camera in the close range, where the markers are most accurate.

Abstract

Robot Localization

Explore similar work

Towards Quadruped-Provided Localization and Active Tracking for Micro-UAVs

Cross-Session 3D LiDAR and Camera Fusion for Robust Localization of Unmanned Aerial Vehicles in GPS-Denied Environments

Learning-Based Dynamic Obstacle Avoidance for a UAV Using Only Three Range Sensors