Counter a drone via deep reinforcement learning
(English) Unmanned aerial vehicles (UAV) also known as drones have been used for a variety of reasons such as surveillance, reconnaissance, shipping and delivery, etc. and commercial drone market growth is expected to reach remarkable levels in the near future. However, drones can accidentally or in...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | España |
| Institución: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/689871 |
| Acceso en línea: | http://hdl.handle.net/10803/689871 https://dx.doi.org/10.5821/dissertation-2117-400823 |
| Access Level: | acceso abierto |
| Palabra clave: | Àrea temàtica UPC: Aeronàutica i espai 629 |
| Sumario: | (English) Unmanned aerial vehicles (UAV) also known as drones have been used for a variety of reasons such as surveillance, reconnaissance, shipping and delivery, etc. and commercial drone market growth is expected to reach remarkable levels in the near future. However, drones can accidentally or intentionally violate the air routes of major airports, flying too close to commercial aircraft or invading the privacy of someone. In order to prevent these unwanted events to happen, counter-drone technology is needed to eliminate the threats coming from drones and hopefully the drones can be integrated into the skies safely. A number of counter-drone solutions are being developed, but the cost of drone detection ground systems can also be very high, depending on the number of sensors deployed and powerful fusion algorithms. Counter-drone system supported by an artificial intelligence (AI) method can be an efficient way to fight against drones instead of human intervention. Considering the recent advances in AI, counter-drone systems with AI can also be very accurate. The time required to engage with the target can be less than other methods based on human intervention such as bringing down a malicious drone by a laser gun. Also, AI can identify and classify the target with a high precision in order to prevent a false interdiction with the targeted object. Counter-drone technology with AI will bring important advantages to the threats coming from some drones and will help the skies to become safer and more secure. AI has been used in different research areas in aerospace to create an intelligent system. Especially, a drone can be controlled by AI methods such as deep reinforcement learning (DRL) in different purposes. With the support of DRL, drones can become more intelligent and eventually they can be fully autonomous. The main objective of this PhD thesis is to develop an artificial intelligence approach based on deep reinforcement learning to counter drones that may pose a threat to safety or security. AI agents can continuously learn and adapt to new threats and countering drones with DRL has several advantages. One of the most important advantages is autonomous decision-making which enables AI agents to make autonomous decisions based on their environment and the situation. In this way, drone threats can be countered quickly and effectively, even in vulnerable environments. Additionally, AI agents can be trained in simulation, allowing for safe experimentation, testing, and validation before deployment. Firstly, DRL architecture is proposed to make drones behave autonomously inside a suburb neighborhood environment. Secondly, a state-of-the-art object detection algorithm for drone detection is also added to the counter drone solution. The construction of drone detection models involves transfer learning and training a state-of-the-art object detection algorithm. After achieving fully autonomous drone which can avoid obstacles in an environment, a deep reinforcement learning method to counter a drone in a 2D space in an environment is presented. In this way, drone can maintain its current altitude, and it can try to catch another drone without crashing any obstacle in the environment. Finally, a deep reinforcement learning model is developed to counter a drone in a challenging 3D space in an environment. The learner drone is not only moving in a 2D space but also changing altitudes to eliminate the target drone. It is important to ensure that AI agents are properly trained and validated so that they can make safe and responsible decisions. Without proper testing and validation, there is a risk that AI agents in sensitive areas such as airports or critical infrastructure might perform actions that could be dangerous or violate regulations. As a result, DRL-based counter-drone solutions can be made more practical, efficient, and secure for future use. |
|---|