A Vision-Based End-to-End Reinforcement Learning Framework for Drone Target Tracking

Drones, Journal Year: 2024, Volume and Issue: 8(11), P. 628 - 628

Published: Oct. 30, 2024

Drone target tracking, which involves instructing drone movement to follow a moving target, encounters several challenges: (1) traditional methods need accurate state estimation of both the and target; (2) conventional Proportional–Derivative (PD) controllers require tedious parameter tuning struggle with nonlinear properties; (3) reinforcement learning methods, though promising, rely on drone’s self-state estimation, adding complexity computational load reducing reliability. To address these challenges, this study proposes an innovative model-free end-to-end framework, VTD3 (Vision-Based Twin Delayed Deep Deterministic Policy Gradient), for tracking tasks. This framework focuses controlling while maintaining specific distance. is pure vision-based algorithm integrates YOLOv8 detector, BoT-SORT algorithm, Gradient (TD3) algorithm. It diminishes reliance GPS other sensors simultaneously enhancing capability complex motion trajectories. In simulated environment, we assess performance across four trajectories (triangular, square, sawtooth, square wave, including scenarios occlusions). The experimental results indicate that our proposed substantially outperforms PD in applications. Across various trajectories, demonstrates significant reduction average errors along X-axis Y-axis up 34.35% 45.36%, respectively. Additionally, it achieves notable improvement 66.10% altitude control precision. terms smoothness, markedly enhances metrics, improvements 37.70% jitter 60.64% Jerk RMS. Empirical verify superiority feasibility tracking.

Language: Английский

---

Empirical Trials on Unmanned Agriculture in Open-Field Farming: Ridge Forming

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(18), P. 8155 - 8155

Published: Sept. 11, 2024

The decreasing rural population and migration to urban areas for high-tech opportunities have weakened the agricultural labor force. While data technology has been adopted in protected agriculture, numerous challenges remain field agriculture. In this study, we focus on one of fundamental steps operations, i.e., ridge forming, specifically unmanned We compared performance a conventional tractor with an implement that ridge-forming robot. operation were collected using acquisition system, comparison between results both methods was conducted. Additionally, analyzed linearity autonomous driving expenses associated selected operation. Our findings indicate developed robot forming caused less torque damage achieved more accurate target soil depth, showing distance error only 0.267 m. Furthermore, it eliminated need hiring operator significantly reduced fuel consumption, which accounts 50.81% operational expenses. These suggest operations can be effectively replaced by systems, further research agriculture is warranted.

Language: Английский

---

MULTIDIRECTIONAL USE OF UNMANNED AERIAL VEHICLES IN THE AREA OF SAFETY

Zeszyty Naukowe SGSP, Journal Year: 2024, Volume and Issue: 1(92), P. 219 - 237

Published: Dec. 30, 2024

The paper presents a wide range of possibilities for the use drones in terms human safety, theenvironment and technical facilities. In accordance with concept paper, these areas aredivided into four main sources hazards each cases using anunmanned aircraft are presented. Hazards caused by flooding, environment, fire thoseoccurring on construction sites subject to analysis. aim was markareas hazard grid superimposed map an area specialised drone camera.Based this information, it is possible define risk people property.In case depth river spillway determined according width ofthe channel, affecting safety residents town. For environmentalaspect, surface water table its fields, meadows pastureswere determined. This also important from agricultural point view, including determiningthe extent crop damage. fire, enabled assessment damaged building as result high temperatures, ofa situation plan related building’s structure context continueduse. final site. Here, turn, objective tomap occupational risks those working there and, particular, identify ofdangerous, harmful nuisance factors. All diagrams presented confirmed widespreaduse UAVs diagnosis levels. With technology, easier todiagnose develop preventive measures.

Language: Английский

---