University of Twente Student Theses
As of Friday, 8 August 2025, the current Student Theses repository is no longer available for thesis uploads. A new Student Theses repository will be available starting Friday, 15 August 2025.
Multi-Agent Collision Avoidance Using PPO In Decentralized Reinforcement Learning For Drone Simulated Environment
Lysenko, Volodymyr (2025) Multi-Agent Collision Avoidance Using PPO In Decentralized Reinforcement Learning For Drone Simulated Environment.
|
PDF
1MB |
| Abstract: | The rapid increase of unmanned aerial vehicles (UAVs) in delivery, inspection and emergency-response tasks is pushing air traffic density beyond human supervision capabilities. Ensuring that swarms of autonomous drones remain safe, especially in urban airspace, using collision-avoidance strategies that are decentralized, data-efficient and robust to partial observability. This research investigates how well decentralized Deep Reinforcement Learning (DRL) policies enable multiple drones to reach assigned goals without collisions when each drone perceives only local, sensor-like information. A custom Gymnasium environment was built to simulate 2-D shared airspace with adjustable traffic density and obstacles number. Proximal Policy Optimization (PPO) served as the baseline algorithm for training the policies which will be responsible for the drone’s movement towards the goal and collision avoidance. Trained policies were evaluated on (i) collision-free episode rate, (ii) goal-completion rate, (iii) training sample efficiency, (iv) average steps for episode completion. An optional experimental branch will extend basic state vectors with simulated Light Detection and Ranging (LiDAR)-style observations to quantify the impact of richer perception on learning speed and final performance. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science, 58 process technology |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/107272 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page