Accurate 3D object detection is crucial for autonomous vehicles (AVs) to navigate safely in complex environments. This paper introduces a novel fusion framework that integrates Camera image-based 2D object detection using YOLOv8 and LiDAR data-based 3D object detection using PointPillars, hence named C2L3-Fusion. Unlike conventional fusion approaches, which often struggle with feature misalignment, C2L3-Fusion enhances spatial consistency and multi-level feature aggregation, significantly improving detection accuracy. Our method outperforms state-of-the-art approaches such as YoPi-CLOCs Fusion Network, standalone YOLOv8, and standalone PointPillars, achieving mean Average Precision (mAP) scores of 89.91% (easy), 79.26% (moderate), and 78.01% (hard) on the KITTI dataset. Successfully implemented on the Nvidia Jetson AGX Xavier embedded platform, C2L3-Fusion maintains real-time performance while enhancing robustness, making it highly suitable for self-driving vehicles. This paper details the methodology, mathematical formulations, algorithmic advancements, and real-world testing of C2L3-Fusion, offering a comprehensive solution for 3D object detection in autonomous navigation.

This paper presents the development of a system for simulating and visualizing autonomous vehicle navigation using a single-beam LiDAR sensor integrated with an artificial neural network, implemented entirely in JavaScript. The single-beam LiDAR sensor in the simulation was configured to closely resemble real-world counterparts, such as the YDLIDAR TG30. The proposed neural network architecture, consisting of four output neurons, processes data from ten distinct angular measurements of the LiDAR sensor to predict necessary steering and speed adjustments for autonomous navigation. Furthermore, we introduce an enhanced collision detection algorithm that improves upon traditional methods by providing more precise identification of line segment intersections. Implementing the entire system in JavaScript demonstrates the viability of utilizing web technologies for complex machine learning tasks, representing a significant advancement in autonomous vehicle research. Experimental results validate the effectiveness of this approach across various scenarios, emphasizing its potential for broader applications in cost-sensitive autonomous systems. This work may also serve as a valuable resource for researchers new to machine learning, contributing to greater accessibility in the development of autonomous vehicles.

Research on integrating camera and LiDAR in self-driving car systems has important scientific significance in the context of developing 4.0 technology and applying artificial intelligence. The research contributes to improving the accuracy in recognizing and locating objects in complex environments. This is an important foundation for further research on optimizing response time and improving the safety of self-driving systems. This study proposes a real-time multi-sensor data fusion method, termed "Multi-Layer Fusion," for object detection and localization in autonomous vehicles. The fusion process leverages pixel-level and feature-level integration, ensuring seamless data synchronization and robust performance under adverse conditions. Experiments conducted on the CARLA simulator. The results show that the method significantly improves environmental perception and object localization, achieving a mean detection accuracy of 95% and a mean distance error of 0.54 meters across diverse conditions, with real-time performance at 30 FPS. These results demonstrate its robustness in both ideal and adverse scenarios

Autonomous mobile robots (AMRs) operating in real-world environments such as warehouses and hospitals require accurate environmental perception for safe navigation. This paper addresses the calibration of heterogeneous sensors – a Red-Green-Blue (RGB) USB camera and an Ouster OS1 Light Detection and Ranging (LiDAR) – and their fusion on a differential-drive AMR running under Robot Operating System version 1 (ROS1) Noetic. The core problem is the absence of pre-recorded extrinsic transforms and the impracticality of live calibration when only offline recorded data are available, rendering standard ROS-based workflows inapplicable. To solve this, a complete three-step pipeline is proposed: 1) offline camera intrinsic calibration from extracted .bag frames using a planar checkerboard method with automated farthest-point sampling for frame diversity; 2) interactive LiDAR-to-camera alignment via axis-mapping selection among 11 rotation candidates followed by fine-tuning with real-time visual feedback; 3) calibration verification by projecting the full LiDAR point cloud onto synchronized camera frames with depth-based colorization. Experiments on a 205-second, 10.5 GB .bag file demonstrate a reprojection error of 0.2748 px (RMS) – well below the 0.5 px excellence threshold – and an average of 7,000–7,500 LiDAR points projected per frame, reaching 12,428 points in optimal scenes. The low error is explained by diversity-based frame selection, which eliminates redundant frames and ensures broad spatial coverage. The pipeline is implemented entirely in Python using standard open-source libraries without GPU, making it suitable for resource-constrained onboard computers in practical robotic deployments.