Autonomous Navigation of Drones

ABSTRACT This work proposes a PrimalDual Neural Network (PDNN) based Robust Model Predictive Control (RMPC) framework to address trajectory tracking challenges for quadrotor unmanned aerial vehicles (UAVs) under unmodeled dynamics, external disturbances, and nonlinearities. Firstly, polytopic uncertain model capturing system uncertainties disturbances is established. A cascade control strategy then introduced decouple underactuated dynamics mitigate nonlinear coupling effects. An RMPC design with modeluncertainty compensation reject the uncertainties. The compensationbased problem formulated as quadratic programming (QP) problem, PDNN optimization method developed ensure fast computation solutions realtime control. inputtostate stability (ISS) of closedloop PDNNRMPC rigorously proven. Experimental results demonstrate that proposed achieves reduction in time compared interiorpoint method. Furthermore, varying existing tubebased MPC Recurrent (RNN)MPC, respectively improves average 4.23% 3.25% position accuracy, by 38.70% 39.17% attitude accuracy. These... Read More

Source

One of the applications unmanned aerial vehicles (UAVs) is to transport various cargoes such as medical supplies, food, and electronic devices. Hence, a significant challenge in cargo transportation mitigating disturbances along trajectory, including external factors like wind uncertainties affecting system model. This study introduces robust adaptive integral fast terminal sliding mode control strategy (AIFTSMC) for an uncertain quadrotor UAV carrying load, faced with adverse influences. The systems integrated model derived using LagrangeEuler method quadrotors translational subsystem NewtonEuler rotational subsystem. Nonlinear methods exhibit promising potential guiding stabilizing this intricate underactuated system, (SMC) techniques standing out their distinctive attributes. Designing IFTSMC technique enables trajectory tracking load management reduce steady-state errors within finite-time frame. By incorporating rules upper-bound estimates undesirable factors, controller consistency enhanced. proposed demonstrates performance finite time. Simulation comparisons exist... Read More

Source

This paper addresses the team orienteering problem applied to unmanned aerial vehicles (UAVs), considering obstacle avoidance and environmental factors such as wind conditions payload weight. The objective is optimize UAV routes maximize collected rewards while adhering operational constraints. To achieve this, we employ a simheuristic algorithm for overall route optimization, integrating A* determine feasible paths between nodes that avoid obstacles in 2D grid-based environment. Then, feedforward neural network estimates travel time based on speed, conditions, trajectory distance, estimation incorporated into optimization process improve planning accuracy. Numerical experiments evaluate impact of various parameters, including placement, These include maps with 30 100 points interest varying densities show our hybrid method improves solution quality by up 15% total profit compared baseline approach. Furthermore, computation times remain within 510% baseline, showing added predictive layer maintains computational efficiency.

Source

Unmanned aerial vehicle (UAV) self-localization in complex environments is critical when global navigation satellite systems (GNSSs) are unreliable. Existing datasets, often limited to low-altitude urban scenes, hinder generalization. This study introduces Multi-UAV, a novel dataset with 17.4 k high-resolution UAVsatellite image pairs from diverse terrains (urban, rural, mountainous, farmland, coastal) and altitudes across China, enhancing cross-view geolocalization research. We propose lightweight value reduction pyramid transformer (VRPT) for efficient feature extraction residual network (RFPN) multi-scale fusion. Using meter-level accuracy (MA@K) relative distance score (RDS), VRPT achieves robust, high-precision localization varied terrains, offering significant potential resource-constrained UAV deployment.

Source

Automated flight response for unmanned aerial vehicles (UAVs) to comply with airspace restrictions around airports and other no-fly zones. The UAV calculates its distance from restricted areas using its own location and the location of the restrictions. If the distance is below a threshold, it takes immediate action like landing or preventing takeoff. If the distance is greater, it allows normal flight. This allows automated compliance with airspace rules while providing flexibility for safe operations when farther away.

Source

Obstacle avoidance for drones that balances proactive and reactive collision avoidance while tracking targets. The drone dynamically adjusts its flight characteristics based on obstacle proximity. If an obstacle is far away, it proactively moves to maintain a safe distance. If an obstacle is close, it reacts quickly to avoid collision. The drone evaluates candidate motion adjustments and selects the best one based on route optimization. This allows it to proactively steer around obstacles that may cause immediate threats versus reactively evading imminent collisions.

Source

Autonomous landing of drones in complex environments using adaptive path planning and deep reinforcement learning. The method selects between local and global path optimization based on perception range. For local optimization, drones plan paths around nearby obstacles. For global optimization, they use perceived frontiers. This improves efficiency by avoiding redundant planning. For landing, a neural network learns to control the drone using reward functions. This increases neural network update efficiency compared to traditional methods.

Source

This article studies an anti-collision control method for the trajectory tracking problem of unmanned aerial vehicle aircraft skin inspection under complex wind disturbance. To guarantee safety during inspection, disturbance is described by maximum position offset constraint vehicles (UAVs). Then, exponential nonlinear integral super-twisting sliding mode (ENISTSM) controller designed, and incorporated with prescribed performance control(PPC) to ensure that error consistently constrained within specified bounds. Subsequently, attitude angular velocity cascaded law investigated based on modified Rodrigues parameters (MRPs) representation using (ESTSM) method. The proposed achieves fast relatively large variable angles robust In addition, stability controllers proven via Lyapunov analysis. Finally, simulation results are included considering turbulent field demonstrate effectiveness advantages.

Source

Robust guidance and control for unmanned vehicles that allows autonomous navigation to waypoints while ensuring stability and robustness. The approach involves integrated guidance and feedback control that generates trajectories for both position and attitude. The position trajectory is tracked using a force along the vehicle's thrust direction. The attitude trajectory is generated based on the desired thrust direction. This allows finite-time stable attitude control. The overall system is discretized in time for computerized integration. The discretized dynamics and control are proven to have almost global asymptotic stability.

Source

Development platform for autonomous vehicles like drones that enables custom applications and skills using high-level behavioral objectives. The platform provides APIs, SDKs, and tools to let developers build apps that control the autonomous vehicle by specifying intuitive, high-level intentions instead of directly piloting. The objectives are executed by the vehicle's autonomous system. The platform hides the underlying complexity of autonomous navigation. It supports applications like automatic return to home, object tracking, and environmental perception. The objectives can be learned from user selections to guide automatic selection in flight. The platform also provides simulation environments for developing and testing applications.

Source

Safe landing point search method for unmanned aerial vehicles (UAVs) on unfamiliar terrain using terrain maps and contour lines. The method involves generating contour lines based on a minimum height and interval from a terrain map. Then, it searches for largest empty circles (LECs) with a minimum radius in the contour map. The UAV is provided the LEC with the largest radius as a safe landing spot. This leverages contour lines to find flattest areas without obstructions for UAV landing.

Source

In recent years, drone technology has seen profound advancements, especially with regards to safe and autonomous operation, which heavily relies on object detection avoidance capabilities. These autonomously functioning drones can operate in challenging environments for tasks like search rescue operations, as well industrial monitoring. The present research focuses enhancing by utilizing publicly available image datasets instead of custom images. Datasets VisDrone, DroneDeploy, DOTA contain a plethora stunning, real-life images that make them ideal candidates improving the accuracy robustness models. We propose an optimized method training YOLOv5 model enhance detection. collected dataset undergoes evaluation from precision, recall, F1-score, mAP through both CNN YOLO findings show using deep architecture implement real-time UAVs is more efficient than traditional approaches.

Source

This work develops an improved integrated planning and control framework for unmanned aerial vehicle (UAV) in complex environments with dense obstacles to achieve fast accurate path planning, trajectory generation, tracking control. Utilizing the potential function-based rapid-exploration random tree star (P-RRT*), a bidirectional dynamic informed P-RRT* (BDIP-RRT*) algorithm is first introduced enhance sampling efficiency, facilitating swift generation. To further optimize initial path, greedy employed minimize redundant segments within generated path. Subsequently, points are assigned based on original using adaptive distance interpolation strategy. A hybrid optimized generator considering jerk snap built obtain reference UAV. Moreover, two prescribed-time laws designed ensure UAV position attitude Finally, simulation results performed illustrate effectiveness superior performances of developed scheme.

Source

Optimizing energy usage and engine life of hybrid electric aircraft engines by intelligently managing battery power. The technique involves analyzing flight data and engine health to determine if using battery power during flight instead of ground taxiing will increase engine life or prevent exceeding limits. If so, e-taxiing is skipped and battery power is applied to assist the engine spools during flight. This extends time on wing and avoids exceeding limits, but can also affect component durability. The decision balances energy savings vs. component life tradeoffs.

Source

Turbomachine rotor design with improved blade retention to prevent blade loss during flight. The rotor has a movable ring with teeth that engage hooks on the blade roots. A protective element, like a coating or fixed part, is provided on the inner faces of the teeth. This prevents contact wear and fretting between the hooks and teeth, which can deform and wear the blades. The protective element reduces blade damage from impacts and vibrations during flight.

Source

Beaded composite structural web for aircraft, particularly rotorcraft, that provides weight savings and impact resistance compared to conventional aluminum webs. The beaded composite web is a single layer composite panel with integrated stiffening beads on the exterior surface. The beads prevent buckling and allow flexing during impacts to absorb energy and reduce damage compared to rigid webs. The beaded composite web is lighter than equivalent aluminum webs with internal stiffeners. It can be used for components like rotorcraft torque boxes to improve impact resistance and weight.

Source

Automated system that optimizes robot movement instructions to reduce wear and prolong robot lifespan. The system continuously monitors motor current during robot operation and compares it to secondary currents. By analyzing the current differences, it identifies inefficiencies in the robot's command instructions that cause excessive torque and acceleration. The system then generates modified instructions to address those inefficiencies, reducing motor load and wear.

Source

Landing zone detection of autonomous aerial vehicles is crucial for locating suitable landing areas. Currently, localization predominantly relies on methods that use RGB cameras. These sensors offer the advantage integration into majority vehicles. However, they lack depth perception, which can lead to suggestion non-viable zones, as only assess an area using information. They do not consider if surface irregular or accessible a user (easily person foot). An alternative approach utilize 3D information extracted from images, but this introduces challenge correctly interpreting ambiguity. Motivated by latter, we propose methodology segmentation DNN-Superpixel approach. This consists three steps: First, proposal involves clustering superpixels segment, locate, and delimit zones within scene. Second, feature extraction adjacent objects through bounding box analyzed area. Finally, uses Deep Neural Network (DNN) segment landable non-landable, considering its accessibility. The experimental results are feasible promising. For example, achieved average recall 0.953, meaning identified 95.3% ... Read More

Source

Quadcopter drones are capable of executing complex aerobatic maneuvers when controlled manually by skilled pilots but limited to simple actions flying autonomously in open spaces. As such, this study introduces a comprehensive system that enables generate and execute sophisticated environments with dense obstacle distributions. A universal representation is proposed, succinctly capturing flight as series discrete intentions. These intentions consist topology attitude changes, which can be combined various ways describe intricate maneuvers. spatial-temporal joint optimization trajectory planner also introduced dynamically feasible trajectories smooth possible devoid collisions. In addition, we investigate unique yaw sensitivity issues identify the inherent influence differential flatness singularities on rotations while avoiding associated dynamics issues. ablation studies confirmed necessity these compensation strategies. Additional simulations physical experiments validated stability feasibility our proposed for improving uncrewed aerial flight. The achieve performance usually reser... Read More

Source

Helicoidal composite materials with improved impact resistance and damage tolerance. The materials have a unique layered structure that spirals around the part, rather than being flat. This helical architecture allows for more design freedom and tailoring of the composite properties. The helical layup can be made using thin ply unidirectional (TPUD) fabric, thin ply woven fabric (TPW), or quasi-unidirectional woven fabric (QUDW). The helical layup provides better impact resistance compared to traditional flat layups because it allows for more controlled fiber orientation and delamination prevention.

Source