Autonomous Flights inside Narrow Tunnels

TL;DR

This work tackles autonomous navigation of multirotors through narrow tunnels by integrating a virtual omni-directional perception module with a perception-and-disturbance-aware planner. It introduces on-board RGB-D sensing from three directions, EDF-based mapping, and a multi-front-end state estimator to maintain robust localization in feature-poor environments. The planning framework couples perception quality and ego airflow disturbances into trajectory, yaw, and speed optimization, including cross-section recognition and active yaw strategies to maintain safe centerline following. Extensive real-world experiments across straight, 2-D, and 3-D tunnels—including vents and construction-site pipes—demonstrate the system's ability to traverse complex, arbitrarily oriented tunnels as narrow as $0.5$ m, outperforming an experienced pilot and a prior method, with an extendable pipeline for deployment on other multirotor platforms.

Abstract

Multirotors are usually desired to enter confined narrow tunnels that are barely accessible to humans in various applications including inspection, search and rescue, and so on. This task is extremely challenging since the lack of geometric features and illuminations, together with the limited field of view, cause problems in perception; the restricted space and significant ego airflow disturbances induce control issues. This paper introduces an autonomous aerial system designed for navigation through tunnels as narrow as 0.5 m in diameter. The real-time and online system includes a virtual omni-directional perception module tailored for the mission and a novel motion planner that incorporates perception and ego airflow disturbance factors modeled using camera projections and computational fluid dynamics analyses, respectively. Extensive flight experiments on a custom-designed quadrotor are conducted in multiple realistic narrow tunnels to validate the superior performance of the system, even over human pilots, proving its potential for real applications. Additionally, a deployment pipeline on other multirotor platforms is outlined and open-source packages are provided for future developments.

Figures (35)

• Figure 1: The narrow tunnels to test the proposed autonomous aerial system.
• Figure 2: The customized quadrotor platform for narrow tunnel flights.
• Figure 3: The system architecture and the workflow for narrow tunnel flights. The perception module adopts the color and depth images, as well as IMU data to produce RGBD-inertial odometry for state estimation and then performs map fusion. The tunnel center waypoints extracted from the mapping result facilitate the perception-and-disturbance-aware planning, which generates flight trajectories, yaw trajectories and speed profiles. The results are then utilized to generate control commands and executed by the motors. The workflow consists of pre-tunnel initialization, which perform pre-planning to the localized tunnel entrance according to the detected marker, intra-tunnel peri-replanning and ex-tunnel post-planning.
• Figure 4: The illustration of entrance localization of the tunnel shown in Fig. \ref{['fig:tunnel_2d']}. The color code indicates the height; the axis indicates the pose of the quadrotor and the black arrow indicates the estimated tunnel entrance pose.
• Figure 5: The system architecture of the state estimator in narrow tunnels. The estimator processes color and depth images from individual camera modules to produce RGBD-inertial odometries. It includes multiple parallel front ends, a front end coordinator, and an optimization back-end. In practice, it utilizes three RGB-D camera modules.