Sapience Competition 1

Where It All Began: Autonomy Without GPS, Indoors

The first Sapience competition challenged us with a deceptively simple question: can a team of aerial robots explore an unknown indoor environment, entirely on their own, and come back with information a first responder could actually use? Indoors there is no GPS, no motion-capture system, and no second chance — everything the drones know about the world they have to perceive, estimate, and decide on board, in real time.

Leading a student team at the University of Klagenfurt, we built our first complete autonomy stack for exactly this setting: aerial robots that enter an unknown building, localize themselves purely from onboard sensing, map their surroundings, and search for objects of interest — the way a rescue team would sweep a collapsed or hazardous structure.

Technical Overview

The indoor setting forced us to solve the hard problems first:

• GPS-denied localization — visual-inertial odometry as the primary state estimate, designed to survive low-texture corridors and abrupt lighting changes.
• Real-time onboard mapping — incremental 3D reconstruction of the environment during flight, used simultaneously for navigation and for the mission output.
• Autonomous exploration — frontier-based exploration to systematically cover unknown space without human input.
• Semantic perception — onboard object detection to identify and annotate items of interest on the map as the search progressed.

All computation ran on board the vehicles — no external computers, no motion capture, no communication lifeline.

From Simulation to Real-World Demonstrations

We developed and validated the full stack in simulation before moving to hardware, then iterated relentlessly in field tests until the system flew reliably in cluttered, previously unseen indoor spaces. The gap between a system that works in simulation and one that survives a real building is where most of this project’s lessons were learned — sensor timing, degraded perception, and recovery behaviors when the environment refuses to cooperate.

Broader Impact and Future Directions

Competition 1 laid the technical and organizational foundation for everything that followed: the modular autonomy architecture, the sensor-fusion approach, and the team workflow were carried forward and extended into the outdoor setting of Sapience Competition 2 — and ultimately into our research group’s contribution to larger initiatives on fully autonomous flight.

Reflections on Teamwork and Leadership

Beyond the technical aspects, this project has been a formative experience in leading interdisciplinary collaboration under real-world constraints. From aligning hardware-software interfaces to defining mission strategies and resolving failures in the field, the journey demanded constant adaptation, clear communication, and a shared vision within the team.

I would like to thank the brilliant students who made this possible:

• Gilbert Tanner
• Georg Steinthaler
• Tim Schumann
• Ben Wesse
• Jonas Spieler

Your dedication, creativity, and resilience shaped this project from the initial concept to a system ready for real-world deployment.