Sapience Competition 2 | Luca Di Pierno

From Concept to Competition: Leading a Student Project Toward Real-World Impact

Over the course of the past 1.5 years, I have led a student team at the University of Klagenfurt in an ambitious research-driven competition centered on a highly relevant topic in robotics: multi-agent autonomous systems for search and rescue operations in environments where GPS is not available. This challenge lies at the intersection of robotic autonomy, real-time perception, and collaborative decision-making—domains that are becoming increasingly critical in modern disaster response scenarios.

The core objective of our project was to design and implement an integrated system of aerial robots capable of autonomously exploring unknown environments, constructing semantic maps, and identifying actionable insights such as hazardous zones, safe access routes, or potential victims. These functionalities are intended to support first responders like the Red Cross during critical interventions in complex and potentially dangerous conditions.

Left: one of our UAV platforms, fully equipped for onboard perception and computation. Middle: presenting the system architecture during the competition briefing. Right: our drones in flight during an autonomous mission run.

Field tests for final integration testing, at the Hobby Airport in Klagenfurt, Austria.

Technical Overview

Our system architecture was designed from the ground up with modularity, adaptability, and robustness in mind. It features:

Real-time perception and semantic awareness, including object detection, dynamic tracking, and the generation of annotated maps.
Multi-layered planning and execution, with both global and local path planners integrated with obstacle avoidance strategies and dynamic path adaptation.
A modular autonomy stack, built around our own finite-state-machine and behavior-tree-based architecture, enabling flexible decision-making and reactivity in unstructured environments.
Sensor fusion and redundancy, allowing the system to dynamically switch between multiple odometry and localization sources to ensure reliable navigation even in degraded conditions.

The entire system was deployed on board our UAVs and executed fully autonomously, with no external computation or communication infrastructure.

Presenting our autonomy stack to the judges during the system presentation — and preparing one of our UAVs for its next fully autonomous flight.

From Simulation to Real-World Demonstrations

What makes this project particularly rewarding is its real-world validation across diverse scenarios. The system was not only tested in the structured environment of competitions but also deployed in vastly different field settings, without requiring major architectural changes. Among the successful demonstrations were:

Autonomous exploration and mapping of a highway tunnel, including obstacle avoidance and semantic annotation of detected elements.
3D reconstruction of a freeway bridge, generated using onboard visual and Lidar-based mapping tools during an autonomous flight.
Exploration of complex indoor buildings, with semantic detection and real-time annotation of objects of interest, simulating real-world search and rescue missions.

These tests not only demonstrated the maturity and robustness of our architecture, but also proved its adaptability to different environments and mission profiles.

The Competition in Alabama

The campaign culminated at the international competition hosted at the University of Alabama in Huntsville in July 2025, where our system had to prove itself head-to-head against teams from around the world — under real field conditions, with fully autonomous missions and no room for manual intervention.

Our team after winning the international competition at the University of Alabama in Huntsville, July 2025 — and one of our UAVs during a competition run.

Broader Impact and Future Directions

The insights gained from this project—and the resulting software and system design—have contributed directly to our involvement in larger, high-impact initiatives such as SPRIND – Fully Autonomous Flights 2.0, funded by the German Federal Agency for Disruptive Innovation. Within this project, our research group (Control of Networked Systems) is actively contributing toward the development of fully autonomous aerial platforms for complex missions, building upon the core principles established in our earlier work.

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