Deep-Reinforcement-Learning-Based Collision Avoidance in UAV Environment

Abstract

Unmanned Aerial Vehicles (UAVs) have recently attracted both academia and industry representatives due to their utilization in tremendous emerging applications. Most UAV applications adopt Visual Line of Sight (VLOS) due to ongoing regulations. There is a consensus between industry for extending UAVs' commercial operations to cover the urban and populated area controlled airspace Beyond VLOS (BVLOS). There is ongoing regulation for enabling BVLOS UAV management. Regrettably, this comes with unavoidable challenges related to UAVs' autonomy for detecting and avoiding static and mobile objects. An intelligent component should either be deployed onboard the UAV or at a Multi-Access Edge Computing (MEC) that can read the gathered data from different UAV's sensors, process them, and then make the right decision to detect and avoid the physical collision. The sensing data should be collected using various sensors but not limited to Lidar, depth camera, video, or ultrasonic. This paper proposes probabilistic and Deep Reinforcement Learning (DRL)-based algorithms for avoiding collisions while saving energy consumption. The proposed algorithms can be either run on top of the UAV or at the MEC according to the UAV capacity and the task overhead. We have designed and developed our algorithms to work for any environment without a need for any prior knowledge. The proposed solutions have been evaluated in a harsh environment that consists of many UAVs moving randomly in a small area without any correlation. The obtained results demonstrated the efficiency of these solutions for avoiding the collision while saving energy consumption in familiar and unfamiliar environments.

Review of Sensor Technology to Support Automated Air-to-Air Refueling of a Probe Configured Uncrewed Aircraft

“…When operating multiple UAs with crossing flight paths, collision avoidance systems become a critical safety component for safe and efficient operations. In terms of perception, EO/IR systems and tracking algorithms have been researched to assist active sensors, such as sonar or radar, in providing perception for a collision avoidance system to ensure safe deconfliction [ 74 , 75 ]. Understanding that these EO/IR systems provide perception for avoidance, these systems could also use the same perception information to inform a GNC on how to join and maintain formation with another aircraft [ 76 ].…”

Deep Reinforcement Learning for Internet of Drones Networks: Issues and Research Directions

“…Drones utilize the data acquired by their different sensors, such as Lidar, depth camera, video, or ultrasonic, to achieve that task. Unlike the work in [37], the authors in [38] propose probabilistic and DQN-based algorithms to prevent collisions while minimizing energy consumption in IoD networks considering limited knowledge about the environment. Their introduced technique can be run on board the drone or at a multi-access edge computing entity, according to the drone capacity and the task overhead.…”

Path Planning Research of a UAV Base Station Searching for Disaster Victims’ Location Information Based on Deep Reinforcement Learning

“…The UAV air base station has the characteristics of strong maneuverability, controllable mobility, and convenient deployment and can support the high-speed transmission of communication data, etc. [ 3 ]. The application of UAV base stations in disaster relief scenarios improves the problem of communication signals being difficult to reach in complex environments, and also makes it possible to provide all-round coverage of communication signals in disaster relief mission areas.…”