Teaching Drones to Navigate Autonomously with Artificial Intelligence

15/09/2026

Researchers from the Information Processing and Telecommunications Center (IPTC) at Universidad Politécnica de Madrid have developed a platform to train and validate autonomous drone navigation using Reinforcement Learning (RL).

The platform uses the AirSim simulator to create realistic flight scenarios involving multiple Unmanned Aerial Vehicles (UAVs). Instead of programming every possible situation in advance, Reinforcement Learning allows a drone to learn how to make decisions by interacting with its environment and receiving rewards or penalties depending on its actions. The researchers validated the platform using two UAVs and neural networks trained with the Soft Actor-Critic (SAC) algorithm.

One of the key contributions is the possibility of creating different training, validation and testing scenarios, helping researchers assess whether the learned behaviour works reliably in different situations. The modular architecture also makes it easier to adapt the platform to different types of drones and control systems.

This research can contribute to the development of more autonomous and intelligent drones, with potential applications in areas such as surveillance, infrastructure inspection, emergency response and other situations where drones need to navigate dynamic environments with limited human intervention. The platform also provides a useful environment for studying machine learning in systems where reliability and safety are particularly important.

This research provides a valuable step towards the development of safer and more reliable autonomous UAV systems. By enabling navigation algorithms to be trained and evaluated in realistic and reproducible simulated environments, the platform can help bridge the gap between artificial intelligence research and future real-world applications, including infrastructure inspection, emergency response, surveillance and coordinated multi-drone operations.

Bibliographic reference:

Álvarez-Novoa, D., Pérez-Muñoz, A.G., Alonso, A., & Pérez, M.S. A platform for the navigation of Unmanned Aerial Vehicles based on Reinforcement Learning in Jornadas de Automática, 47, 6 pp. https://doi.org/10.17979/ja-cea.2026.47.13789

Authors:

Ángel Grover Pérez Muñoz: GS / ORCID / LinkedIn

Alejandro Alonso Muñoz: GS / ORCID / LinkedIn

LinkedIn: https://www.linkedin.com/company/iptc-upm/

For more information: www.iptc.upm.es