Remote Autonomy for Multiple Small Lowcost UAVs in GNSS-denied Search and Rescue Operations

TL;DR

This work tackles autonomous flight in GNSS-denied disaster scenarios using inexpensive, unmodified DJI drones. It introduces a remote-autonomy pipeline with an on-board Android app for state estimation and obstacle avoidance and a ground-control station that fuses multi-UAV perception into a globally consistent 3D map, enabling a single operator to supervise a fleet. Key contributions include a ROS-based GCS for multi-UAV fusion, an on-device VIO adapted for GNSS-denied state estimation, reactive obstacle avoidance via $P$-field planning, and a door-traversal mode for cluttered passages, validated through latency, waypoint accuracy, obstacle avoidance, door traversal, and 3D mapping evaluations. The approach promises scalable, cost-efficient SAR operations by reducing pilot workload and permitting rapid, coordinated reconnaissance in challenging environments.

Abstract

In recent years, consumer-grade UAVs have been widely adopted by first responders. In general, they are operated manually, which requires trained pilots, especially in unknown GNSS-denied environments and in the vicinity of structures. Autonomous flight can facilitate the application of UAVs and reduce operator strain. However, autonomous systems usually require special programming interfaces, custom sensor setups, and strong onboard computers, which limits a broader deployment. We present a system for autonomous flight using lightweight consumer-grade DJI drones. They are controlled by an Android app for state estimation and obstacle avoidance directly running on the UAV's remote control. Our ground control station enables a single operator to configure and supervise multiple heterogeneous UAVs at once. Furthermore, it combines the observations of all UAVs into a joint 3D environment model for improved situational awareness.

Figures (8)

• Figure 1: Multiple consumer-grade UAVs supervised by a single operator facilitate Search & Rescue in an industrial hall. Left: External view of the UAVs; Right: Created 3D RGB map with UAV trajectories (colored arrows).
• Figure 2: Remote autonomy system architecture. Software modules are divided into two layers: For each UAV, the safety-relevant modules are executed on the UAV's remote control unit (RCU). It consists of an Android device and a remote controller (RC), connected via the DJI MSDK. A WiFi Protobuf server connects each RCU to the ground control station (GCS), where a single operator configures and supervises multiple UAVs and where globally consistent 3D environment models are created and visualized. UAVs from other manufactures can be connected over a MAVROS interface.
• Figure 3: User interfaces on our remote controller (RCU, a) and at the ground control station (GCS, b). RCU: Various status indicators are overlayed on the video feed for the safety pilot including UAV heading and virtual stick commands. The example shows the thermal image and tracked VIO features on the color image of a DJI M3T in side-by-side mode. GCS: The fleet management shows the current UAVs' states including control options (b, bottom left). The current waypoint mission is shown in the joint 3D environment model (b, bottom right) and overlayed on a satellite image (b, top left). Live-streams (b, top right) allow direct supervision and targeted inspection.
• Figure 4: Door detection within horizontal obstacle scan. The current UAV pose is marked by the coordinate axes. The green line indicates a door candidate that is filtered out since its normal deviates too much from the scan's radial direction. The blue line represents the selected door candidate. The red and blue arrows correspond to the planned pre- and post-traversal poses, respectively.
• Figure 5: Evaluating position control accuracy. Horizontal black lines depict five target positions with the gray-shaded area indicating to the target threshold. The colored lines show the trajectories of five different runs as estimated by the VIO.