We live in the ‘era’ of Unmanned Aircraft Systems (UAS), an all-encompassing term that includes the aircraft or the UAV, the ground-based controller (the person or software agent operating the machine), and the communications systems connecting the two ^[1]. Furthermore, new solutions are being proposed, allowing one to interconnect the different components of this system using cloud-oriented architecture ^[2].

Today, the UAVs are revolutionizing the world and businesses in a way that hardly anyone could have ever imagined ^[3][4]. This rapid evolution of UAVs and more importantly their commercial application in an ever-larger spectrum of scenarios, increases the need for safer and more reliable solutions ^[5][6][7][8]. At the same time, the incremental developments of sensing technologies (e.g., thermal, multispectral, and hyperspectral) in multiple areas may change parts of society by creating new solutions and applications ^{[9][10][11][12][13]}.

In order to safely deploy UAVs in an urban environment, UAVs must reach the same level of reliability and safety as cars ^[14]. This level of safety and reliability must be assured regardless of the operation conditions and occurrence of unexpected events. This leads to the obvious conclusion that, just as for other autonomous vehicles, it is required to develop a collision avoidance architecture that is agnostic of the environment constrains, and is capable of finding solutions to unexpected events in real-time. The lack of such robust architectures has led to multiple disasters in the past ^{[15][16][17][18][19][20][21]}, which will naturally increase with the expected exponential growth in the number of UAVs that will be deployed (both by consumers and companies). In this video, a robust architecture, part of a complete Framework for Fully Autonomous UAVs (FFAU) is proposed along with complete implementation (both software and hardware) that satisfy the functional and technical requirements of the proposed architecture. The increased safety of the proposed framework allows new and interesting usage scenarios such as the long distance remote control of UAVs. Furthermore, a novel Dynamic Collision Avoidance algorithm is proposed, which utilizes a Convolutional Neural Network (CNN) to extract features from video frames, and a Recursive Neural Network (RNN) that takes advantage of the video temporal characteristics capable of estimating if there is an incoming collision, such as the one depicted in Figure 1. This figure illustrates two frames from a video, retrieved from the ColANet dataset ^[22], that presents a kid kicking a ball into a UAV.

Figure 1. Novel Dynamic Collision Avoidance Detection. The images represent two frames from the same video, where a boy kicks a ball into a UAV.

There are quite a few attempts to integrate UAVs into an architecture that links different UAVs modules and a cloud platform. The architecture presented in ^[23] covers three main topics: Air traffic control network, cellular network, and the Internet. The layered architecture provides services for different UAV applications, such as surveillance or search and rescue. The video does not present any implementation of the proposed architecture and only outlines general concepts of the IoD (Internet-of-Drones), which later had a partial implementation ^[2]. Neither papers gave much importance to a key factor: The UAV itself. Contrarily, this video analyzes in greater depth the UAV functional blocks without disregarding the equally important cloud connection.