Unmanned Aerial Vehicles (UAVs) are increasingly being used in many challenging and diversified applications. These applications belong to the civilian and the military fields. To name a few; infrastructure inspection, traffic patrolling, remote sensing, mapping, surveillance, rescuing humans and animals, environment monitoring, and Intelligence, Surveillance, Target Acquisition, and Reconnaissance (ISTAR) operations. However, the use of UAVs in these applications needs a substantial level of autonomy. In other words, UAVs should have the ability to accomplish planned missions in unexpected situations without requiring human intervention. To ensure this level of autonomy, many artificial intelligence algorithms were designed. These algorithms targeted the guidance, navigation, and control (GNC) of UAVs. In this paper, we described the state of the art of one subset of these algorithms: the deep reinforcement learning (DRL) techniques. We made a detailed description of them, and we deduced the current limitations in this area. We noted that most of these DRL methods were designed to ensure stable and smooth UAV navigation by training computer-simulated environments. We realized that further research efforts are needed to address the challenges that restrain their deployment in real-life scenarios.

Most applications for small unmanned aerial vehicles (sUAVs) have a critical need for appropriate position and attitude control in environments with extreme external dynamic disturbances such as wind gusts. Moreover, to maximize flight time, UAVs have to operate within strict constraints on payload and under computational efficiency. This article presents an optimized robust adaptive controller for a UAV in the presence of realistic wind gusts that are flying in an urban environment at an altitude lower than 400 ft. The position controller is based on two degree-of-freedom proportional-integral-derivative controller tuned with particle swarm optimization, while the attitude controll is based on a robust adaptive integral backstepping controller. By assuming the knowledge of the predetermined limits of the external and unstructured disruptions (for example, wind gusts) at low altitude, a guaranteed quality of transient and steady-state tracking performance were attained. The aerodynamics, wind gust model, and control modules are integrated into a six-degree-of-freedom UAV with a fully nonlinear robust adaptive controller. Optimal power is obtained for UAVs using the radial inflow model in the presence of wind gusts that are compared with the simplified model. Two case studies were performed systematically for two representative flight paths, namely, up-cruise-down and circular paths. The simulation results demonstrate the robustness and adaptive property of controllers against wind gusts. These results are useful for a variety of UAV applications, e.g. accurate trajectory tracking and autonomous waypoint navigation without loss of performance in the presence of wind disturbances under computational efficacy.

This article presents an optimal tracking controller retrofitted with a nonlinear adaptive integral compensator, specifically designed to ensure robust and accurate positioning of Vertical Take-Off and Landing (VTOL) Unmanned Aerial Vehicles (UAVs) that utilize contra-rotating motorized propellers for differential thrust generation. The baseline position controller is synthesized by employing a fixed-gain Linear Quadratic Integral (LQI) tracking controller that stabilizes position by tracking both state variations and pitch-axis tracking error integral, which adjusts the voltage to control each coaxial propeller’s speed accurately. Additionally, the baseline tracking control law is supplemented with a rate-varying integral compensator. It operates as a nonlinear scaling function of the tracking-error velocity and the braking acceleration to enhance the accuracy of reference tracking without sacrificing its robustness against exogenous disruptions. The controller’s performance is analyzed by performing experiments on a tailored hardware-in-the-loop aero-pendulum testbed, which is representative of VTOL UAV dynamics. Experimental results demonstrate significant improvements over the nominal LQI tracking controller, achieving 17.9%, 61.6%, 83.4%, 43.7%, 35.8%, and 6.8% enhancement in root mean squared error, settling time, overshoot during start-up, overshoot under impulsive disturbance, disturbance recovery time, and control energy expenditure, respectively, underscoring the controller’s effectiveness for potential UAV and drone applications under exogenous disturbances.

This paper addresses the time-varying rendezvous problem of a swarm of drones with the leader-follower consensus hierarchy. Each drone in the swarm is under the influence of external disturbances in the form of wind gusts. To control the swarm of perturbed drones, this paper proposes a fully distributed adaptive leader-following time-varying disturbance rejection pinning control for the rendezvous of drones using the framework of the extended state observer that depends on the state estimates of the neighboring agents rather than the actual states. The communication topology within the swarm is modelled using algebraic graph theory. Unlike most of the literature in which the leader agent act as a virtual reference generator, an active leader agent with non-zero control input is considered. The states of the leader drone are only available to a subset of the graph. The communication topology between leader-follower is directed while that between follower-follower is undirected. Based on the relative output information between the neighboring drones, the extended state observer estimates the local states and disturbances of each drone and the adaptive control actively compensates for the external disturbances simultaneously. Finally, it is shown analytically as well as in simulations that the time-varying rendezvous can be achieved by the proposed algorithm effectively.

The development of autonomous, fast, agile small Unmanned Aerial Vehicles (UAVs) brings up fundamental challenges in dynamic environments with fast and agile maneuvers, unreliable state estimation, imperfect sensing, coupling action, and perception in real-time under severe resource constraints. However, autonomous drone racing is a challenging research problem at the intersection of computer vision, planning, state estimation, and control. To bridge this, we propose an approach in the context of autonomous, perception-action aware vision-based drone racing in a photorealistic environment. Our approach integrates a deep convolutional neural network (CNN) with state-of-the-art path planning, state estimation, and control algorithms. The developed deep learning method is based on computer vision approaches to detecting the gates and estimating the flyable area. The planner and controller then use this information to generate a short, minimum-snap trajectory segment and send corresponding motor commands to reach the desired goal. A thorough evaluation of our proposed methodology has been carried out using the Gazebo and FlightGoggles (photorealistic sensor) environments. Extensive experiments demonstrate that the proposed approach outperforms state-of-the-art methods and flies the drone more consistently than many human pilots. Moreover, we demonstrated that our proposed system successfully guided the drone through tight race courses, reaching speeds up to 7m/s of the 2019 AlphaPilot Challenge.