Cost-Effective Autonomous Drone Navigation Using Reinforcement Learning: Simulation and Real-World Validation DOI Creative Commons

Tomasz Czarnecki,

Marek Stawowy,

Adam Kadłubowski

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 15(1), P. 179 - 179

Published: Dec. 28, 2024

Artificial intelligence (AI) is used in tasks that usually require human intelligence. The motivation behind this study the growing interest deploying AI public spaces, particularly autonomous vehicles such as flying drones, to address challenges navigation and control. primary challenge lies developing a robust, cost-effective system capable of real-world environments, handling obstacles, adapting dynamic conditions. To tackle this, we propose novel approach integrating machine learning (ML) algorithms, specifically, reinforcement (RL), with comprehensive simulation testing framework. Reinforcement algorithms designed solve problems requiring optimization solution for highest possible reward were used. It was assumed do not have be created from scratch, but they need well-defined training environment will appropriately or punish actions taken. This aims develop implement drone using algorithms. innovation integration ML control system, encompassing both simulations testing. A vital component creating multi-stage accurately replicates actual flight conditions progressively increases complexity scenarios, ensuring robust evaluation algorithm performance. research also introduces new optimizing cost accessibility. involves commercially available, drones open-source free tools, significantly reducing entry barriers potential users. critical aspect assess whether affordable components can provide sufficient accuracy stability without compromising quality. authors developed autonomously determining optimal paths controlling drone, allowing it avoid obstacles respond real time. performance trained confirmed through flights, which allowed assessing their usefulness practical scenarios.

Language: Английский

Developing a model for managing sustainable woody biomass resources in the Fuji region of Japanese temperate climate: Reinforcement learning-based optimization DOI
Weiheng Wang, Hyun Bae Kim, Takuyuki Yoshioka

et al.

Biomass and Bioenergy, Journal Year: 2025, Volume and Issue: 200, P. 107972 - 107972

Published: May 20, 2025

Language: Английский

Citations

0
Fuzzy Reinforcement Learning Algorithm for Efficient Task Scheduling in Fog-Cloud IoT-Based Systems DOI
Reyhane Ghafari, N. Mansouri

Journal of Grid Computing, Journal Year: 2024, Volume and Issue: 22(4)

Published: Sept. 23, 2024

Language: Английский

Citations

2
Generalization Enhancement of Visual Reinforcement Learning through Internal States DOI Creative Commons
Hanlin Yang, William Zhu, Xianchao Zhu

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(14), P. 4513 - 4513

Published: July 12, 2024

Visual reinforcement learning is important in various practical applications, such as video games, robotic manipulation, and autonomous navigation. However, a major challenge visual the generalization to unseen environments, that is, how agents manage environments with previously backgrounds. This issue triggered mainly by high unpredictability inherent high-dimensional observation space. To deal this problem, techniques including domain randomization data augmentation have been explored; nevertheless, these methods still cannot attain satisfactory result. paper proposes new method named Internal States Simulation Auxiliary (ISSA), which uses internal states improve tasks. Our contains two agents, teacher agent student agent: has ability directly access environment’s used facilitate agent’s training; receives initial guidance from subsequently continues learn independently. From another perspective, our can be divided into phases, transfer phase traditional phase. In first phase, interacts imparts knowledge vision-based agent. With of agent, able discover more effective representations address next autonomously learns information environment, ultimately, it becomes enhanced generalization. The effectiveness evaluated using DMControl Generalization Benchmark DrawerWorld texture distortions. Preliminary results indicate significantly improves performance complex continuous control

Language: Английский

Citations

0
Cooperative-Competitive Decision-Making in Resource Management: A Reinforcement Learning Perspective DOI
Artem Isakov, Danil Peregorodiev,

Pavel Brunko

et al.

Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 375 - 386

Published: Nov. 14, 2024

Language: Английский

Citations

0
Reinforcement Learning-Based Control for Collaborative Robotic Brain Retraction DOI Creative Commons

Ibai Inziarte-Hidalgo,

Estela Nieto, Diego Roldán

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(24), P. 8150 - 8150

Published: Dec. 20, 2024

In recent years, the application of AI has expanded rapidly across various fields. However, it faced challenges in establishing a foothold medicine, particularly invasive medical procedures. Medical algorithms and devices must meet strict regulatory standards before they can be approved for use on humans. Additionally, robots are often custom-built, leading to high costs. This paper introduces cost-effective brain retraction robot designed perform The is trained, specifically Deep Deterministic Policy Gradient (DDPG) algorithm, using reinforcement learning techniques with contact model, offering more affordable solution such delicate tasks.

Language: Английский

Citations

0
Advanced Multi-Robot Path Planning and Control Architecture for Precision Cancer Treatment Systems DOI Open Access

Sridevi Palepu

International Journal of Scientific Research in Computer Science Engineering and Information Technology, Journal Year: 2024, Volume and Issue: 10(6), P. 1957 - 1964

Published: Dec. 15, 2024

This article presents an innovative approach to mobile robot path planning and control systems specifically designed for cancer detection treatment applications in medical environments. introduces a novel prioritized path-planning algorithm that enables multiple robots navigate collision-free while maintaining precise coordination during procedures. The system architecture integrates advanced technologies, including ATL COM/VC++ components, digital/analog interfacing, COM/.NET interoperable objects with C# user controls XML comprehensive machine management. incorporates fuzzy logic learning techniques intelligent collision avoidance, alongside artificial neural networks generative AI models pattern classification forecasting. implementation leverages communication protocols, TCP/IP, RS232, CAN, USB, ensure robust connectivity across all components. Extensive testing through black/white box methodologies, regression testing, simulation of pneumatic, hydraulic, PLC components demonstrates the system's reliability precision. shows significant improvements efficiency, response times, overall compared existing solutions. suggests this integrated not only enhances accuracy procedures but also provides scalable framework future robotics applications. successful validation clinical settings indicates its potential widespread adoption facilities, marking substantial advancement automated robotics.

Language: Английский

Citations

0
Predicting outcomes using neural networks in the intensive care unit DOI
GR Sridhar,

Venkat Yarabati,

Lakshmi Gumpeny

et al.

World Journal of Clinical Cases, Journal Year: 2024, Volume and Issue: 13(11)

Published: Dec. 25, 2024

Patients in intensive care units (ICUs) require rapid critical decision making. Modern ICUs are data rich, where information streams from diverse sources. Machine learning (ML) and neural networks (NN) can leverage the rich for prognostication clinical care. They handle complex nonlinear relationships medical have advantages over traditional predictive methods. A number of models used: (1) Feedforward networks; (2) Recurrent NN convolutional to predict key outcomes such as mortality, length stay ICU likelihood complications. Current exist silos; their integration into workflow requires greater transparency on that analyzed. Most accurate enough use operate 'black-boxes' which logic behind making is opaque. Advances occurred see through opacity peer processing black-box. In near future ML positioned help far beyond what currently possible. Transparency first step toward validation followed by trust adoption. summary, NNs transformative ability enhance accuracy improve patient management ICUs. The concept should soon be turning reality.

Language: Английский

Citations

0
Cost-Effective Autonomous Drone Navigation Using Reinforcement Learning: Simulation and Real-World Validation DOI Creative Commons

Tomasz Czarnecki,

Marek Stawowy,

Adam Kadłubowski

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 15(1), P. 179 - 179

Published: Dec. 28, 2024

Artificial intelligence (AI) is used in tasks that usually require human intelligence. The motivation behind this study the growing interest deploying AI public spaces, particularly autonomous vehicles such as flying drones, to address challenges navigation and control. primary challenge lies developing a robust, cost-effective system capable of real-world environments, handling obstacles, adapting dynamic conditions. To tackle this, we propose novel approach integrating machine learning (ML) algorithms, specifically, reinforcement (RL), with comprehensive simulation testing framework. Reinforcement algorithms designed solve problems requiring optimization solution for highest possible reward were used. It was assumed do not have be created from scratch, but they need well-defined training environment will appropriately or punish actions taken. This aims develop implement drone using algorithms. innovation integration ML control system, encompassing both simulations testing. A vital component creating multi-stage accurately replicates actual flight conditions progressively increases complexity scenarios, ensuring robust evaluation algorithm performance. research also introduces new optimizing cost accessibility. involves commercially available, drones open-source free tools, significantly reducing entry barriers potential users. critical aspect assess whether affordable components can provide sufficient accuracy stability without compromising quality. authors developed autonomously determining optimal paths controlling drone, allowing it avoid obstacles respond real time. performance trained confirmed through flights, which allowed assessing their usefulness practical scenarios.

Language: Английский

Citations

0