Abstract
Precise, real-time 3D localization is a prerequisite for the coordinated flight of Micro Aerial Vehicle (MAV) swarms. While Ultra-Wideband (UWB) radio ranging provides a low-cost indoor alternative to expensive motion capture systems, it often suffers from significant latency and low update rates. This paper introduces a robust localization system that fuses UWB ranging with Inertial Measurement Unit (IMU) data using an Extended Kalman Filter (EKF). By augmenting the state vector to estimate acceleration bias, the proposed method achieves an 80Hz update rate with near-zero latency, significantly improving flight stability and accuracy compared to UWB-only approaches.
Introduction
The deployment of MAV swarms in indoor environments or temporary venues necessitates a localization solution that is both accurate and scalable. Global Positioning System (GPS) signals are typically unavailable or imprecise in these settings. While optical motion capture systems like VICON provide centimeter-level ground truth, their high cost and infrastructure requirements limit their use in large-scale or field deployments.
Ultra-Wideband (UWB) technology has emerged as a promising alternative due to its high-accuracy ranging capabilities. However, standard UWB localization algorithms often rely on constant velocity assumptions, leading to significant estimation lag (up to 500ms) during aggressive maneuvers. This latency can cause flight controllers to diverge, resulting in instability or collisions within a swarm.
Methodology
The core of the proposed system is an Extended Kalman Filter (EKF) designed to fuse absolute UWB position measurements with high-frequency IMU motion sensing.
Augmented State Vector and Control Input
To address the drift inherent in low-cost IMUs, the researchers augmented the EKF state vector to include acceleration bias ($a_{bx}, a_{by}, a_{bz}$) alongside position and velocity. In this architecture, raw IMU acceleration is treated as the control input ($u_{k-1}$) for the prediction step, rather than a standard measurement. This allows the filter to "learn" and compensate for sensor errors in real-time.
Robustness and Hardware
To mitigate "multipath" interference—where UWB signals reflect off surfaces or other drones—the system employs innovation thresholding. Measurements are discarded if the difference between predicted and measured distances exceeds a predefined threshold (e.g., 2 meters).
The hardware setup utilizes TimeDomain UWB sensors integrated with a Pixhawk-based flight controller. For swarm operations, a network of six UWB "anchor" nodes is deployed to establish a common coordinate frame for the mobile nodes mounted on the MAVs.
Experimental Evaluation and Results
The system was evaluated through single-MAV flights compared against VICON ground truth and a real-world 6-MAV swarm performance at the Changi Exhibition Centre in Singapore.
Performance Comparison
The Fusion EKF demonstrated a marked improvement over the standard UWB-only (Vanilla EKF) approach across all critical metrics:
- Latency: The most significant gain was the reduction of latency from approximately 500ms to nearly zero.
- Accuracy: Mean localization error was reduced from 0.30m to 0.16m, with maximum error dropping from 0.71m to 0.39m.
- Update Rate: The system maintained a consistent 80Hz localization frequency, providing the high-bandwidth feedback required for stable autonomous flight.
Discussion and Key Insights
The experimental results highlight that for dynamic flight, latency is often a more critical factor than raw static accuracy. The "Vanilla" EKF's 500ms delay is sufficient to destabilize a drone's flight controller. By incorporating acceleration bias estimation, the Fusion EKF provides a smooth, zero-latency estimate that remains robust even when UWB signals are temporarily occluded by other drones in a dense swarm.
Limitations and Future Work
Despite its success, the system faces challenges during high-G events, such as landing impacts, where massive acceleration spikes can temporarily degrade EKF performance. Additionally, the geometry of UWB anchor placement significantly influences triangulation accuracy. Future research will focus on optimizing UWB airtime and communication protocols to scale the system for swarms containing dozens or hundreds of MAVs.
Summary updated