cs.ROApr 9, 2026

AgiPIX: Bridging Simulation and Reality in Indoor Aerial Inspection

Authors: Sasanka Kuruppu Arachchige, Juan Jose Garcia, Changda Tian, Lauri Suomela, Panos Trahanias, Adriana Tapus, Joni-Kristian Kämäräinen

Organizations: Computing Sciences department at Tampere University, Finland · U2IS at ENSTA, Paris · Institute of Computer Science, Foundation for Research and Technology - Hellas (FORTH), Heraklion, Greece.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Autonomous indoor flight for critical asset inspection presents fundamental challenges in perception, planning, control, and learning. Despite rapid progress, there is still a lack of a compact, active-sensing, open-source platform that is reproducible across simulation and real-world operation. To address this gap, we present Agipix, a co-designed open hardware and software platform for indoor aerial autonomy and critical asset inspection. Agipix features a compact, hardware-synchronized active-sensing platform with onboard GPU-accelerated compute that is capable of agile flight; a containerized ROS~2-based modular autonomy stack; and a photorealistic digital twin of the hardware platform together with a reliable UI. These elements enable rapid iteration via zero-shot transfer of containerized autonomy components between simulation and real flights. We demonstrate trajectory tracking and exploration performance using onboard sensing in industrial indoor environments. All hardware designs, simulation assets, and containerized software are released openly together with documentation.

Figures & tables

Fig. 2 : Comparison of consumer and research platforms by onboard compute, agility, and sensing. Criteria include openness, ROS 2 support, simulation (double ticks denote photorealism), UI, CPU performance ( https://www.cpubenchmark.net ), and GPU availability. Agility is measured by thrust-to-weight; diagonal span is in flight configuration. (*conditionally open-source)
Fig. 3 : System overview of the AgiPIX stack. AgiAUTO runs modular ROS 2 components on the companion computer. Flight-state and setpoint topics are bridged to PX4 via native DDS in both simulation and real flights. An external ROS 2 logger records system telemetry. AgiUI provides operator interaction, logging, and mission commands over a MQTT [ 36 ] .
ComponentProductSpecification
FrameCustom open-source4 mm4\text{\,}\mathrm{mm} carbon fiber
MotorT-Motor Slatts 230623 ×\times 6 mm6\text{\,}\mathrm{mm} stator, 2400 kV2400\text{\,}\mathrm{kV} , 758 W758\text{\,}\mathrm{W}
PropellerAzure Power SFP51485.1 i5.1\text{\,}\mathrm{i} nch length and 4.8 i4.8\text{\,}\mathrm{i} nch pitch
BatteryTattoo G-Tech 45006 ×\times 3.7 V3.7\text{\,}\mathrm{V} , 4500 mA h4500\text{\,}\mathrm{mA}\text{\,}\mathrm{h}
Flight ControllerPixhawk OrangeBuilt in redundancy
Motor ControllerF55A Pro II 3-6SDShot protocol, 4 ×\times 60 A60\text{\,}\mathrm{A}
TABLE I : Overview of the components of the flight hardware design.
Fig. 4 : Hardware time synchronization pipeline. The Pixhawk (IMU) provides the master clock; the LiDAR is synchronized via PPS, with an ESP32 generating the required timing signal.
Fig. 5 : This is the complete overview of the AgiUI where the user can see different perspectives of the robots it is controlling, including a top view, a side view, and a third-person view.
TrajectoryAgiSIM ATE (m)AgiREAL ATE (m)
Lemniscate0.06370.1404
Up Down Spiral0.03680.1049
TABLE II : Trajectory-tracking performance comparison between AgiSIM and AgiREAL across different trajectories.
Fig. 6 : Trajectory-tracking comparison of Ground Truth, AgiSIM, and AgiREAL in two trajectories at 6 m/s maximum velocity; Up down spiral and lemniscate: (1) spiral in the xx – yy plane, (2) spiral in the yy – zz plane, and (3) lemniscate in the xx – yy plane.
Fig. 7 : Left: AKW Zwentendorf, Right: AgiPIXduring the various stages of the trial.
Fig. 8 : Representative ENRICH 2025 mapping output.
Fig. 9 : Percentage of free, occupied, and unmapped volume over the total exploration time for the exploration mode using the closest-frontier method for the given map
Fig. 10 : Drone trajectory relative to the mapping trajectory when controlled by FAINT [ 27 ] , a learning-based visual navigation method. The drone successfully reached the goal on the left trajectory, but missed it on the right.
Software ComponentCPU Utilisation (%)GPU Utilisation (%)
AdaptiveLIO135
Livox Driver92
Map Manager420
Ego planner110
Autonomous flight100
Controller10
TABLE III : Average system utilisation on Jetson Orin NX Super 16 GB by AgiAUTO.
Fig. 11 : AgiPIX supports heterogeneous hardware deployments. AgiREAL::UAV and AgiREAL::UGV, shown here, were deployed in ENRICH 2025 [ 46 ] .
More in this topic

Aerial Robotics

Aerial robotics studies the design, perception, control, and coordination of flying robots for navigation, inspection, transport, and physical interaction.

See all 340 papers

Explore similar work

CardsList

  1. A Modular Dual-Camera Pipeline for Micro-Inspection Using Aerial Robots

    Jun 9, 2026S. H. Mirtajadini, N. Rublein, R. M. Ramakrishnan +3Aerial RoboticsRobotic Inspection

    Smart Mechatronics and Robotics (SMART) Research Group, Saxion University of Applied Sciences, Enschede, The Netherlands · Faculty of Science and Engineering, University of Groningen, The Netherlands

  2. AirSimAG: A High-Fidelity Simulation Platform for Air-Ground Collaborative Robotics

    Mar 24, 2026Yangjie Cui, Xin Dong, Boyang Gao +3Robotics SimulationRobotics

    School of Aeronautic Science and Engineering, Beihang University, Beijing 100191, China · Hangzhou International Innovation Institute, Beihang University, Hangzhou 311115, China · Institution of Unmanned System, Beihang University, Beijing 100191, China

  3. CARLA-Air: Fly Drones Inside a CARLA World -- A Unified Infrastructure for Air-Ground Embodied Intelligence

    Mar 30, 2026Tianle Zeng, Yanci Wen, Hong ZhangRobotics SimulationAutonomous Driving Simulation

    1Shenzhen Key Laboratory of Robotics and Computer Vision, Southern University of Science and Technology.