Navigation System For Autonomous Vehicle Research Articles

VisioSignNet: A Dual-Interactive Neural Network for enhanced traffic sign detection

Effective traffic sign detection is crucial for the safety and operational efficiency of autonomous vehicle navigation systems, particularly in dynamically changing environments. Addressing the primary challenges of long-range pixel dependencies and enhancing the detectability of small objects in complex scenes, we present VisioSignNet: A Dual-Interactive Neural Network designed for enhanced traffic sign detection. This architecture incorporates Local and Global Interactive Modules (LGIM) and Enhancing Channel and Space Interaction (ECSI) modules. The LGIM is engineered to balance local and global feature interactions, while the ECSI optimizes the interchange of information across channel and spatial dimensions. Their synergistic interaction not only enhances the perceptual field at early processing stages but also significantly improves the recognition of small-scale, critical traffic signs. Evaluated on the TT100K and GTSDB datasets, VisioSignNet achieved mean average precision (mAP) scores of 90.5% and 97%, respectively, with a model size of 26M parameters. Its enhanced variant, VisioSignNet_l, with 34M parameters, reached mAP scores of 93.2% and 97.8%. These outcomes substantiate VisioSignNet’s efficacy in tackling the complexities of traffic sign detection, confirming its potential as a robust solution in the field of autonomous driving technologies.

Deep Q-Learning-Based Trajectory Optimization for Vehicle Navigation in CARLA

This paper presents a comprehensive study that focuses on simulating a vehicle within the autonomousvehicle simulator, CARLA. The primary objective of this research is to enable the vehicle to accurately follow apredetermined trajectory while effectively avoiding obstacles in its environment. Deep Q-Learning algorithms areemployed to achieve this goal, aiming to optimize the safety of the vehicle'snavigation. The simulation of the vehicleserves as a platform for studying the rules of Deep Q-Networks (DQN) and their impact on the vehicle's navigation.The objective is to identify the most suitable rule that leads to improved optimization of the vehicle's trajectory. Byleveraging the capabilities of CARLA as the simulation environment and implementing state-of-the-art DQNalgorithms, this research contributes to the advancement of autonomous vehicle technology. The findings of this studyhave practical implications for enhancing the safety and efficiency of autonomous vehicle navigation systems, makingthem highly relevant to industryprofessionals, researchers, and academic scholars in this field.

EVALUATING NAVIGATION PERFORMANCE OF ELASTICALLY CONSTRUCTED HD MAP WITH MULTI-SENSOR FUSION ENGINE SYSTEM

Abstract. In response to the rapid development of autonomous vehicles and the increasing demand for HD maps, the conventional mapping processes following HD maps guidelines require significant manpower and time resources. Therefore, we propose flexible procedures and methods for HD maps creation, aiming to reduce cost expenditure by employing diverse source of ground control point, sensor data collection, and mapping algorithm. This approach accelerates the production speed and capability of HD maps. In this study, we select Taiwan's National Highway No. 8 as the trial field for the elastic HD map construction method, and equipped with autonomous vehicle-grade GNSS, IMU, and LiDAR systems.We align the constructed-map data to the global coordinate system, in order to realize the concept of control point cloud map. To assess the assistance and correction capabilities of HD maps in autonomous vehicle navigation systems, we conduct accuracy evaluation through both direct and indirect methods, and analyse the strengths and weaknesses of each approach. The analysis result demonstrates that the elastic method-built HD maps not only meet the mapping accuracy requirements specified in the HD maps verification and validation guidelines, but also assist autonomous vehicles in realizing positioning, navigation, and timing with “where in lane” level (0.5 meter) accuracy.

Investigation of Accuracy of TOA and SNR of Radio Pulsar Signals for Vehicles Navigation.

It is known that X-ray and gamma-ray pulsars can only be observed by spacecraft because signals from these pulsars are impossible to be detected on the Earth's surface due to their strong absorption by the Earth's atmosphere. The article is devoted to the theoretical aspects regarding the development of an autonomous radio navigation system for transport with a small receiving antenna, using radio signals from pulsars, similar to navigation systems for space navigation. Like GNSS systems (X-ray and radio), they use signals from four suitable pulsars to position the object. These radio pulsars (out of 50) are not uniformly distributed but are grouped in certain directions (at least 6 clusters can be determined). When using small antennas (with an area of up to tens of square meters) for pulsar navigation, the energy of the pulsar signals received within a few minutes is extremely insufficient to obtain the required level of SNR at the output of the receiver to form TOA estimation, ensuring positioning accuracy up to tens of kilometers. This is one of the scientific tasks that is solved in the paper by studying the relationship between the SNR of the receiver output, which depends on the size of the antenna, the type of signal processing, and the magnitude of the TOA accuracy estimate. The second scientific task that is solved in the paper is the adaptation of all the possible approaches and algorithms suggested in the statistical theory of radars in the suggested signal algorithm for antenna processing and to evaluate the parameters of the TOA and DS pulsar signals, in order to increase the SNR ratio at the receiver output, while preserving the dimensions of the antenna. In this paper, the functional structure of signal processing in a pulsar transport navigation system is proposed, and the choice of the observed second and millisecond pulsars for obtaining a more accurate TOA estimate is discussed. The proposed estimates of positioning accuracy (TOA only, no phase) in an autonomous pulsar vehicle navigation system would only be suitable for the navigation of large vehicles (sea, air, or land) that do not require accurate navigation at sea, air, or desert. Large-sized antennas with an area of tens of square meters to hundreds of square meters can be installed in such vehicles.

A Curbside Parking Planning and Searching Navigation System for Autonomous Vehicles

A Curbside Parking Planning and Searching Navigation System for Autonomous Vehicles

ANALYSIS OF AUTOMATIC CONTROL SYSTEMS OF A MULTIPURPOSE AUTONOMOUS UNMANNED UNDERWATER VEHICLE WITH COMPLEX MOTION DYNAMICS

The overview article is devoted to the analysis of control systems of autonomous unmanned underwater vehicles, which belong to robotic underwater complexes and are used to provide search and deep-sea emergency and rescue operations, survey of various underwater objects (main pipelines, bottom structures, ship hulls, etc.), conducting scientific oceanographic surveys research, monitoring of the marine environment. The work examines scientific literary sources, which provide the results of development and research of orientation, navigation and motion control systems of multi-purpose autonomous unmanned underwater vehicles with complex movement dynamics, in particular when performing missions in a limited space with movement along complex trajectories. Special attention is focused on such modes of movement of devices as trajectory tracking, path-following and way point tracking. An analysis of the means of navigation and control, which can be used to ensure different modes of movement, was carried out, as well as the directions of development of both the unmanned underwater vehicles themselves and their control systems were considered.
 In order to substantiate the development of a control system for a multi-purpose maneuverable mini-class autonomous vehicle, the main requirements are formulated. It is proposed to conduct a synthesis and research of the control system based on the simulation of the dynamics of the movement of the mini-vehicle for different modes of movement in terms of parameters and trajectories, to develop algorithms for the adaptive control of the vehicle, to improve the system orientation and navigation, built on the basis of MEMS sensors. An information model of the development process is proposed and substantiated, which specifies the sequence of determining important parameters and characteristics of the vehicle and performing important stages of synthesis and analysis of the control system.

Accelerated and Refined Lane-Level Route-Planning Method Based on a New Road Network Model for Autonomous Vehicle Navigation

Lane-level route planning is a critical issue for a lane-level navigation system for autonomous vehicles. Current route-planning methods mainly focus on the road level and applying them directly to search for lane-level routes results in a reduction in search efficiency. In addition, previously developed lane-level methods lack consideration for vehicle characteristics and adaptability to multiple road network structures. To solve this issue, this study proposes an accelerated and refined lane-level route-planning algorithm based on a new lane-level road network model. First, five sub-layers are designed to refine the internal structure of the divided road and intersection areas so that the model can express multiple variations in road network structures. Then, a multi-level route-planning algorithm is designed for sequential planning at the road level, lane group level, lane section level, and lane level to reduce the search space and significantly improve routing efficiency. Last, an optimal lane determination algorithm considering traffic rules, vehicle characteristics, and optimization objectives is developed at the lane level to find the optimal lanes on roads with different configurations, including those with a constant or variable number of lanes while satisfying traffic rules and vehicle characteristics. Tests were performed on simulated road networks and a real road network. The results demonstrate the algorithm’s better adaptability to changing road network structures and vehicle characteristics compared with past hierarchical route planning, and its higher efficiency compared with direct route planning, past hierarchical route planning, and the Apollo route-planning method, which can better support autonomous vehicle navigation.

Stereo matching algorithm for autonomous vehicle navigation using integrated matching cost and non-local aggregation

Stereo matching algorithm plays an important role in an autonomous vehicle navigation system to ensure accurate three-dimensional (3D) information is provided. The disparity map produced by the stereo matching algorithm directly impacts the quality of the 3D information provided to the navigation system. However, the accuracy of the matching algorithm is a challenging part to be solved since it is directly affected by the surrounding environment such as different brightness, less texture surface, and different image pair exposure. In this paper, a new framework of stereo matching algorithm that used the integration of census transform (CT) and sum of absolute difference (SAD) at the matching cost computation step, non-local cost aggregation at the second step, winner take all strategy at the third step, and a median filter at the final step to minimize disparity map error. The results show that the accuracy of the disparity map is improved using the proposed methods after some parameter adjustment. Based on the standard Middlebury and KITTI benchmarking dataset, it shows that the proposed framework produced accurate results compared with other established methods.

A review of path following control strategies for autonomous robotic vehicles: Theory, simulations, and experiments

AbstractThis article presents an in‐depth review of path following control strategies that are applicable to a wide range class of marine, ground, and aerial autonomous robotic vehicles. From a control system standpoint, path following can be formulated as the problem of stabilizing a path following error system that describes the dynamics of position and possibly orientation errors of a vehicle with respect to a path, with the errors defined in an appropriate reference frame. In spite of the large variety of path following methods described in the literature we show that, in principle, most of them can be categorized in two groups: stabilization of the path following error system expressed either in the vehicle's body frame or in a frame attached to a “reference point” moving along the path, such as a Frenet‐Serret (F‐S) frame or a Parallel Transport (P‐T) frame. With this observation, we provide a unified formulation that is simple but general enough to cover many methods available in the literature. We then discuss the advantages and disadvantages of each method, comparing them from the design and implementation standpoint. We further show experimental results of the path following methods obtained from field trials testing with under‐actuated and over‐actuated autonomous marine vehicles. In addition, we introduce open‐source Matlab and Gazebo/ROS simulation toolboxes that are helpful in testing path following methods before their integration in the combined guidance, navigation, and control systems of autonomous vehicles.

Characterization, Detection, and Segmentation of Work-Zone Scenes From Naturalistic Driving Data

This paper elucidates the automatic detection and analysis of work zones (construction zones) in naturalistic roadway images. An underlying motivation is to identify locations that may pose a challenge to advanced driver assistance systems (ADAS) or autonomous vehicle navigation systems. We first present an in-depth characterization of work-zone scenes from a custom data set collected from more than a million miles of naturalistic driving data. We then describe two machine learning algorithms based on the ResNet and U-Net architectures. The first approach works in an image classification framework that classifies an image as a work-zone scene or non-work-zone scene. The second algorithm was developed to identify individual components representing evidence of a work zone (signs, barriers, machines, etc.). These systems achieved an [Formula: see text] score of 0.951 for the classification task and an [Formula: see text] score of 0.611 for the segmentation task. We further demonstrate the viability of our proposed models through saliency map analysis and ablation studies. To our knowledge, this is the first study to consider the detection of work zones in large-scale naturalistic data. The systems demonstrate potential for real-time detection of construction zones using forward-looking cameras mounted on automobiles. Such a system can be incorporated with ADAS to assist drivers in navigating through challenging environments such as construction zones, making those areas safer for commuters. The code is available on our GitHub page: https://github.com/VTTI/Segmentation-and-detection-of-work-zone-scenes .

User Interface and AI Navigation System for Autonomous Vehicle

The issue of vehicle accidents due to human error and behavior (rash driving, drink and drive, etc.) is on rising resulting into fatality and economical loss. In order to overcome these issues, the autonomous vehicle is being explored as a probable solution. Driver will be replaced by AI based autonomous system; The main objective is to convert the manual operated EV into autonomous vehicle. The project work is divided into three main parts i.e., Control System for Vehicle Control, AI (Artificial Intelligence) and UI (User Interface). The main objective of User Interface is to give information regarding vehicles health i.e. battery remaining, charging time, distance to be covered, etc. AI will be use as decision making system or we can say brain of the autonomous vehicle. System Control or Vehicle Control System is to control the vehicle’s steering, acceleration and braking system with the help of PID controller while the vehicle is being operated in autonomous mode. The main purpose is to make an autonomous EV for the passengers to go to their selected destination with a single click by selecting destination on display (UI). So the vehicle control is responsible for the vehicle’s steering i.e., at what desired angle the vehicle need to turn, during acceleration the factors that are considered are; at what speed the vehicle needs to move in an straight path or during a turn and the braking system is operated with the help of LIDAR where if it detects an object at what distance it needs to apply the brakes and when to release the brakes. All the factors and conditions required in developing the autonomous vehicle will be considered, as use cases to improve the accuracy of autonomous vehicle, resulting into achieving desired objective of safe travel.

A numerical and experimental study on the obstacle collision avoidance system using a 2D LiDAR sensor for an autonomous surface vehicle

In recent years, great efforts have been made to develop unmanned navigation systems for Autonomous Surface Vehicles (ASVs), while further improvements are still required for ASVs operating around obstacles. In this circumstance, this paper presents an Obstacle Collision Avoidance Guidance (OCAG) using obstacle detection with a two-dimensional (2D) LiDAR sensor. After verifying the algorithm of the OCAG in a numerical manoeuvring simulation, the OCAS was embedded in a physical model of a Catamaran-type ASV with a 2D LiDAR sensor. When the ASV sails along with a predefined global path, the LiDAR sensor detects the obstacles, the collision-avoidance actions are determined based on multiple factors including the motion and orientation information of the vehicle, the measured distance and direction information of the obstacles, as well as the types of the obstacles. The proposed OCAG showed good collision-avoidance performances with different obstacle types along the path.

Autonavi3at Software Interface to Autonomously Navigate on Urban Roads Using Omnidirectional Vision and a Mobile Robot


 
 
 
 
 
 
 The design of efficient autonomous navigation systems for mobile robots or autonomous vehicles is fundamental to perform the programmed tasks. Basically, two kind of sensors are used in urban road following: LIDAR and cameras. LIDAR sensors are highly accurate but expensive and extra work is needed for human understanding of the point cloud scenes; however, visual content is understood better by human beings, which should be used to develop human-robot interfaces. In this work, a computer vision-based urban road following software tool called AutoNavi3AT for mobile robots and autonomous vehicles is presented. The urban road following scheme proposed in AutoNavi3AT uses vanishing point estimation and tracking on panoramic images to control the mobile robot heading on the urban road. To do that, Gabor filters, region growing, and particle filters were used. In addition, laser range data are also employed for local obstacle avoidance. Quantitative results were achieved using two kind of tests, one uses datasets acquired at the Universidad del Valle campus, and field tests using a Pioneer 3AT mobile robot. As a result, important improvements in the vanishing point estimation of 68.26 % and 61.46 % in average were achieved, which is useful for mobile robots and autonomous vehicles when they are moving on urban roads.
 
 
 
 
 
 

A Seamless Navigation System and Applications for Autonomous Vehicles Using a Tightly Coupled GNSS/UWB/INS/Map Integration Scheme

Seamless positioning systems for complex environments have been a popular focus of research on positioning safety for autonomous vehicles (AVs). In particular, the seamless high-precision positioning of AVs indoors and outdoors still poses considerable challenges and requires continuous, reliable, and high-precision positioning information to guarantee the safety of driving. To obtain effective positioning information, multiconstellation global navigation satellite system (multi-GNSS) real-time kinematics (RTK) and an inertial navigation system (INS) have been widely integrated into AVs. However, integrated multi-GNSS and INS applications cannot provide effective and seamless positioning results for AVs in indoor and outdoor environments due to limited satellite availability, multipath effects, frequent signal blockages, and the lack of GNSS signals indoors. In this contribution, multi-GNSS-tightly coupled (TC) RTK/INS technology is developed to solve the positioning problem for a challenging urban outdoor environment. In addition, ultrawideband (UWB)/INS technology is developed to provide accurate and continuous positioning results in indoor environments, and INS and map information are used to identify and eliminate UWB non-line-of-sight (NLOS) errors. Finally, an improved adaptive robust extended Kalman filter (AREKF) algorithm based on a TC integrated single-frequency multi-GNSS-TC RTK/UWB/INS/map system is studied to provide continuous, reliable, high-precision positioning information to AVs in indoor and outdoor environments. Experimental results show that the proposed scheme is capable of seamlessly guaranteeing the positioning accuracy of AVs in complex indoor and outdoor environments involving many measurement outliers and environmental interference effects.

Deep and Transfer Learning Approaches for Pedestrian Identification and Classification in Autonomous Vehicles

Pedestrian detection is at the core of autonomous road vehicle navigation systems as they allow a vehicle to understand where potential hazards lie in the surrounding area and enable it to act in such a way that avoids traffic-accidents, which may result in individuals being harmed. In this work, a review of the convolutional neural networks (CNN) to tackle pedestrian detection is presented. We further present models based on CNN and transfer learning. The CNN model with the VGG-16 architecture is further optimised using the transfer learning approach. This paper demonstrates that the use of image augmentation on training data can yield varying results. In addition, a pre-processing system that can be used to prepare 3D spatial data obtained via LiDAR sensors is proposed. This pre-processing system is able to identify candidate regions that can be put forward for classification, whether that be 3D classification or a combination of 2D and 3D classifications via sensor fusion. We proposed a number of models based on transfer learning and convolutional neural networks and achieved over 98% accuracy with the adaptive transfer learning model.

Road Curb Detection: A Historical Survey.

Curbs are used as physical markers to delimit roads and to redirect traffic into multiple directions (e.g., islands and roundabouts). Detection of road curbs is a fundamental task for autonomous vehicle navigation in urban environments. Since almost two decades, solutions that use various types of sensors, including vision, Light Detection and Ranging (LiDAR) sensors, among others, have emerged to address the curb detection problem. This survey elaborates on the advances of road curb detection problems, a research field that has grown over the last two decades and continues to be the ground for new theoretical and applied developments. We identify the tasks involved in the road curb detection methods and their applications on autonomous vehicle navigation and advanced driver assistance system (ADAS). Finally, we present an analysis on the similarities and differences of the wide variety of contributions.

A Rigorous Observation Model for the Risley Prism-Based Livox Mid-40 Lidar Sensor.

Modern lidar sensors are continuing to decrease in size, weight, and cost, but the demand for fast, abundant, and high-accuracy lidar observations is only increasing. The Livox Mid-40 lidar sensor was designed for use within sense-and-avoid navigation systems for autonomous vehicles, but has also found adoption within aerial mapping systems. In order to characterize the overall quality of the point clouds from the Mid-40 sensor and enable sensor calibration, a rigorous model of the sensor’s raw observations is needed. This paper presents the development of an angular observation model for the Mid-40 sensor, and its application within an extended Kalman filter that uses the sensor’s data to estimate the model’s operating parameters, systematic errors, and the instantaneous prism rotation angles for the Risley prism optical steering mechanism. The analysis suggests that the Mid-40’s angular observations are more accurate than the specifications provided by the manufacturer. Additionally, it is shown that the prism rotation angles can be used within a planar constrained least-squares adjustment to theoretically improve the accuracy of the angular observations of the Mid-40 sensor.

Double Deep Q-Learning and Faster R-CNN-Based Autonomous Vehicle Navigation and Obstacle Avoidance in Dynamic Environment.

Autonomous vehicle navigation in an unknown dynamic environment is crucial for both supervised- and Reinforcement Learning-based autonomous maneuvering. The cooperative fusion of these two learning approaches has the potential to be an effective mechanism to tackle indefinite environmental dynamics. Most of the state-of-the-art autonomous vehicle navigation systems are trained on a specific mapped model with familiar environmental dynamics. However, this research focuses on the cooperative fusion of supervised and Reinforcement Learning technologies for autonomous navigation of land vehicles in a dynamic and unknown environment. The Faster R-CNN, a supervised learning approach, identifies the ambient environmental obstacles for untroubled maneuver of the autonomous vehicle. Whereas, the training policies of Double Deep Q-Learning, a Reinforcement Learning approach, enable the autonomous agent to learn effective navigation decisions form the dynamic environment. The proposed model is primarily tested in a gaming environment similar to the real-world. It exhibits the overall efficiency and effectiveness in the maneuver of autonomous land vehicles.

SCTM: a self-controlled touring and movement for industrial autonomous vehicle navigation

In recent days, the smart industry concept is becoming more prominent in industrial autonomous vehicle navigation systems. The need for service reliability and industrial management has led to various developing ideas in the industrial environment. Autonomous vehicles are commonly used for logistics and demand-aware migration of goods within the industry. In this article, the self-controlled touring and movement (SCTM) method is proposed to design, plan, and execute autonomous vehicle movements. The goal of this method is to improve the output of vehicle movements under controlled operating errors. For this purpose, the user requirement and outputs are identified and estimated beforehand for providing effective vehicle navigation plans. This error, balancing with the demands and outputs, is performed using classification learning wherein the changes of the vehicle’s movements are frequently updated. Further changes are performed within the completion of the tour plan for retaining operational efficiency of the autonomous vehicle. This method is verified using the following metrics: percent of changes, operating error, delay, balancing factor, and response. The proposed SCTM reduces error by 9.73% and delay by 17.1%, whereas it increases the balancing factor by 8.3% for different tour instances.

Obstacle Detection and Safely Navigate the Autonomous Vehicle from Unexpected Obstacles on the Driving Lane.

Nowadays, autonomous vehicle is an active research area, especially after the emergence of machine vision tasks with deep learning. In such a visual navigation system for autonomous vehicle, the controller captures images and predicts information so that the autonomous vehicle can safely navigate. In this paper, we first introduced small and medium-sized obstacles that were intentionally or unintentionally left on the road, which can pose hazards for both autonomous and human driving situations. Then, we discuss Markov random field (MRF) model by fusing three potentials (gradient potential, curvature prior potential, and depth variance potential) to segment the obstacles and non-obstacles into the hazardous environment. Since the segment of obstacles is done by MRF model, we can predict the information to safely navigate the autonomous vehicle form hazardous environment on the roadway by DNN model. We found that our proposed method can segment the obstacles accuracy from the blended background road and improve the navigation skills of the autonomous vehicle.