Deep-Learning-Based Network for Lane Following in Autonomous Vehicles

Electronics, Journal Year: 2022, Volume and Issue: 11(19), P. 3084 - 3084

Published: Sept. 27, 2022

The research field of autonomous self-driving vehicles has recently become increasingly popular. In addition, motion-planning technology is essential for because it mitigates the prevailing on-road obstacles. Herein, a deep-learning-network-based architecture that was integrated with VGG16 and gated recurrent unit (GRU) applied lane-following on roads. normalized input image fed to three-layer output layer as pattern GRU last layer. Next, processed data were two fully connected layers, dropout added in between each Afterward, evaluate model, steering angle speed from control task predicted parameters. Experiments conducted using dataset Udacity simulator real dataset. results show proposed framework remarkably angles different directions. Furthermore, approach achieved higher mean square errors 0.0230 0.0936 inference times 3–4 3 ms. We also implemented our NVIDIA Jetson embedded platform (Jetson Nano 4 GB) compared GPU’s computational time. revealed system took 45–46 s execute single epoch order predict angle. generates fruitful accurate motion planning driving.

Language: Английский

---

Recent Advancements in Deep Learning Applications and Methods for Autonomous Navigation: A Comprehensive Review

Authorea (Authorea), Journal Year: 2023, Volume and Issue: unknown

Published: June 13, 2023

This review article presents recent advancements in deep learning methodologies and applications for autonomous navigation. It analyzes state-of-the-art frameworks used tasks like signal processing, attitude estimation, obstacle detection, scene perception, path planning. The implementation testing of these approaches are critically evaluated, highlighting their strengths, limitations, areas further development. emphasizes the interdisciplinary nature navigation addresses challenges posed by dynamic complex environments, uncertainty, obstacles. With a particular focus on mobile robots, self-driving cars, unmanned aerial vehicles, space vehicles to underscore importance domains. By synthesizing findings from multiple studies, aims be valuable resource researchers practitioners, contributing advancement novel approaches. Key aspects covered include classification applications, methods, general field, innovations, challenges, limitations associated with learning-based systems. also explores current research trends future directions field. extensive overview, initiated 2020, provides all levels, seasoned experts newcomers. Its main purpose is streamline process identifying, evaluating, interpreting relevant research, ultimately progress development technologies.

Language: Английский

---

U-Net-Based Learning Using Enhanced Lane Detection with Directional Lane Attention Maps for Various Driving Environments

Mathematics, Journal Year: 2024, Volume and Issue: 12(8), P. 1206 - 1206

Published: April 17, 2024

Recent advancements in optical and electronic sensor technologies, coupled with the proliferation of computing devices (such as GPUs), have enabled real-time autonomous driving systems to become a reality. Hence, research algorithmic for advanced driver assistance (ADASs) is rapidly expanding, primary focus on enhancing robust lane detection capabilities ensure safe navigation. Given widespread adoption cameras market, relies heavily image data. Recently, CNN-based methods attracted attention due their effective performance tasks. However, expansion global endeavor achieve reliable has encountered challenges presented by diverse environmental conditions road scenarios. This paper presents an approach that focuses detecting lanes areas traversed vehicles equipped cameras. In proposed method, U-Net based framework employed training, additional lane-related information integrated into four-channel input data format considers characteristics. The fourth channel serves edge map (E-attention map), helping modules more specialized learning regarding lane. Additionally, proposition assign weights loss function during training enhances stability speed process, enabling detection. Through ablation experiments, optimization each parameter efficiency method are demonstrated. Also, comparative analysis existing algorithms shows demonstrates superior performance.

Language: Английский

---

An efficient approach for highway lane detection based on the Hough transform and Kalman filter

Innovative Infrastructure Solutions, Journal Year: 2022, Volume and Issue: 7(5)

Published: July 28, 2022

Language: Английский

---

A Survey of Autonomous Driving from a Deep Learning Perspective

ACM Computing Surveys, Journal Year: 2025, Volume and Issue: unknown

Published: April 18, 2025

Autonomous driving represents a significant advancement in the transportation industry, enhancing vehicle intelligence, optimizing traffic management, and improving user experiences. Central to these innovations is deep learning, which enables systems handle complex data make informed decisions. Our survey explores critical applications of learning autonomous driving, such as perception detection, localization mapping, decision-making control. We investigate specialized techniques, including convolutional neural networks, recurrent self-attention transformers, their variants, among others. These methods are applied within various paradigms—supervised, unsupervised reinforcement learning—to suit specific needs driving. analysis evaluates effectiveness, benefits, limitations technologies, focusing on integration with other intelligent algorithms enhance system performance. Furthermore, we examine architectures systems, analyzing how knowledge information organized from modular, pipeline-based frameworks comprehensive end-to-end models. By presenting an exhaustive overview progressing domain bridging research areas, our aims synthesize diverse threads into unified narrative. This effort not only understanding but also pushes boundaries what achievable this interdisciplinary field.

Language: Английский

---

Real-time traffic sign recognition and autonomous vehicle control system using convolutional neural networks

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: April 25, 2025

Language: Английский

---

An Effective Mixed-Precision Quantization Method for Joint Image Deblurring and Edge Detection

Electronics, Journal Year: 2025, Volume and Issue: 14(9), P. 1767 - 1767

Published: April 27, 2025

Deploying deep neural networks (DNNs) for joint image deblurring and edge detection often faces challenges due to large model size, which restricts practical applicability. Although quantization has emerged as an effective solution this issue, conventional methods frequently struggle optimize the unique characteristics of targeted model. This paper introduces a mixed-precision method that dynamically adjusts precision based on regions input image. High-precision is applied neighborhoods preserve critical details, while low-precision employed in other areas reduce computational overhead. In addition, zero-skipping computation strategy designed deployment, thereby enhancing efficiency when processing sparse feature maps. The experimental results demonstrate proposed significantly outperforms existing accuracy across different neighborhood settings (achieving 97.54% 98.23%) also attaining optimal under both 3 × 5 configurations.

Language: Английский

---

A Comprehensive Literature Review on Modular Approaches to Autonomous Driving: Deep Learning for Road and Racing Scenarios

Smart Cities, Journal Year: 2025, Volume and Issue: 8(3), P. 79 - 79

Published: May 6, 2025

Autonomous driving technology is advancing rapidly, driven by integrating advanced intelligent systems. vehicles typically follow a modular structure, organized into perception, planning, and control components. Unlike previous surveys, which often focus on specific system components or single environments, our review uniquely compares both settings, highlighting how deep learning reinforcement methods address the challenges to each. We present an in-depth analysis of local global planning methods, including integration benchmarks, simulations, real-time platforms. Additionally, we compare various evaluation metrics performance outcomes for current methodologies. Finally, offer insights emerging research directions based latest advancements, providing roadmap future innovation in autonomous driving.

Language: Английский

---

Lane segmentation and center line prediction – a multitask framework for autonomous driving systems

Transportation Planning and Technology, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 22

Published: May 13, 2025

Language: Английский

---

Shallow Detail and Semantic Segmentation Combined Bilateral Network Model for Lane Detection

IEEE Transactions on Intelligent Transportation Systems, Journal Year: 2023, Volume and Issue: 24(8), P. 8617 - 8627

Published: July 13, 2023

Lanes play a very critical and important role in maintaining the orderly operation of road traffic. Therefore, automatic lane detection is has significant potential value. Combining existing methods based on deep learning, this paper proposes new model, Bi-Lanenet, to solve problems that still exist current methods, aiming overcome disadvantages these improve practicality algorithm. For task detection, improves segmentation accuracy traditional lanenet model while ensuring network lighter weight more robust. Then, we propose bilateral recognition semantic details, use random sample consensus (RANSAC) method optimize post-processing process. We conduct experiments TuSimple CULane datasets prove can detect lanes image efficiently with 110 frame-per-second (FPS) accurately at 97.08%.

Language: Английский

---