High-definition Maps Research Articles

Methodology of Spatial Data Acquisition and Development of High-Definition Map for Autonomous Vehicles – Case Study from Wrocław, Poland

Autonomous drive systems are a dynamically developed sector of the automotive industry. The key problem in such technological solutions is to provide a reliable navigation system, which is typically based on high-definition (HD) maps supporting the identification of the position of a maneuvering vehicle. HD maps should include possibly up-to-date and detailed information on traffic lanes and on the traffic rules and regulations on such lanes. An effective development of an HD map should be based on the geodetic measurement methods, which ensure efficient and accurate acquisition of spatial data. This article presents the results of an experiment consisting in the manipulation of data obtained with the use of the mobile laser scanning method and further in employing this data in the development of an HD map in an open-source environment. The applied measurement technology and the processing method allowed data of high resolution (frequently above 1000 points per m2) and of high accuracy (3D accuracy down to less than 5 cm). The obtained data were processed in the Vector Map Builder environment (which is accessible from the level of an internet browser) and the final product - HD map was created in the Lanelet2 open-source environment. The above-described experiments allowed two main conclusions. Most importantly, they demonstrate the importance of planning and performing in-field mobile laser scanning measurements. They also point to the important role of the human analyst who needs to manually vectorize the key elements of road infrastructure and to define traffic rules.

Deep Reinforcement Learning-Based Integrated Control of Hybrid Electric Vehicles Driven by Lane-Level High-Definition Map

For automobiles, the innovation of autopilot, connected communication, and new energy have been treated as the trend. This paper proposed a deep reinforcement learning-based integrated control strategy driven by the lane-level high-definition map for hybrid electric vehicles. Firstly, an optimal route with a length of 1277km from Shanghai to Beijing was determined from Google Maps. Based on the longitude, latitude, and elevation data from Google Earth, a lane-level driving environment was built and the global speed was preplanned based on the limited and safe speed. Secondly, a multi-policy integrated control framework and training method were proposed, adopting two types of deep reinforcement learning algorithms to control the velocity and steering angle of the vehicle layer and power distribution of the powertrain. Finally, processor-in-the-loop tests were completed by adopting NVIDIA Jetson Nano and Jetson AGX Xavier. Results show that the integrated control strategy achieves stable control effects on velocity and steering angle based on the environment model, gets a fuel economy of 6.77L/100km, and presents ideal efficiency on embedded devices. Moreover, the results are compared with three other typical powertrains in long-distance driving, and the comparison of driving costs and infrastructure emphasizes the necessity of the development of hybrid electric vehicles.

An overview of the development of key technologies in autonomous driving technology

Technological advances in each of the smaller fields of automated driving technology have collectively advanced the maturity of automated driving technology. This paper describes some of the technological advances in the development of automated driving of automobiles. Firstly, it introduces the improvement of 3D target detection algorithms, the problem of 3D target detection from LIDAR point cloud, and the proposed control network to address the lack of human driving logic in the existing driverless strategy. Next, the convenience of virtual simulation technology in testing self-driving cars is presented as well as the use of high-definition maps by self-driving cars to understand the road and surroundings. Then the solution to the problem of weak feature extraction and low training efficiency when extracting data in the self-driving learning model, the path planning technology for self-driving cars, and the understanding of the scene when self-driving cars are traveling at night are presented. Finally, image target recognition based on synthetic aperture radar, image color recognition method based on deep learning, and how deep learning can be applied to the field of image recognition are presented.

Data Issues in High-Definition Maps Furniture – A Survey

The rapid advancements in sensing techniques, networking, and artificial intelligence (AI) algorithms in recent years have brought autonomous driving vehicles closer to common use in vehicular transportation. One of the fundamental components to enable autonomous driving functionalities are High-Definition (HD) maps – a type of map that carries highly accurate and much richer information than conventional maps. The creation and use of HD maps rely on advances in multiple disciplines, such as computer vision/object perception, geographic information systems, sensing, simultaneous localization and mapping, machine learning, etc. To date, several survey papers have been published describing the literature related to HD maps and their use in specialized contexts. In this survey, we aim to provide (1) a comprehensive overview of the issues and solutions related to HD maps and their use without attachment to a particular context; (2) a detailed coverage of the important domain knowledge of HD map furniture, from acquisition techniques and extraction approaches, through HD map–related datasets, to furniture quality assessment metrics, for the purpose of providing a comprehensive understanding of the entire workflow of HD map furniture generation, as well as its use.

3D Road Boundary Extraction Based on Machine Learning Strategy Using LiDAR and Image-Derived MMS Point Clouds.

The precise extraction of road boundaries is an essential task to obtain road infrastructure data that can support various applications, such as maintenance, autonomous driving, vehicle navigation, and the generation of high-definition maps (HD map). Despite promising outcomes in prior studies, challenges persist in road extraction, particularly in discerning diverse road types. The proposed methodology integrates state-of-the-art techniques like DBSCAN and RANSAC, aiming to establish a universally applicable approach for diverse mobile mapping systems. This effort represents a pioneering step in extracting road information from image-based point cloud data. To assess the efficacy of the proposed method, we conducted experiments using a large-scale dataset acquired by two mobile mapping systems on the Yıldız Technical University campus; one system was configured as a mobile LiDAR system (MLS), while the other was equipped with cameras to operate as a photogrammetry-based mobile mapping system (MMS). Using manually measured reference road boundary data, we evaluated the completeness, correctness, and quality parameters of the road extraction performance of our proposed method based on two datasets. The completeness rates were 93.2% and 84.5%, while the correctness rates were 98.6% and 93.6%, respectively. The overall quality of the road curb extraction was 93.9% and 84.5% for the two datasets. Our proposed algorithm is capable of accurately extracting straight or curved road boundaries and curbs from complex point cloud data that includes vehicles, pedestrians, and other obstacles in urban environment. Furthermore, our experiments demonstrate that the algorithm can be applied to point cloud data acquired from different systems, such as MLS and MMS, with varying spatial resolutions and accuracy levels.

Vision-HD: road change detection and registration using images and high-definition maps

High-definition (HD) maps are becoming increasingly important for autonomous vehicles and advanced driver assistance systems (ADAS), as they provide detailed and accurate information about the road environment. Vision localization using HD maps can improve accuracy, but maps can become outdated. Image perception can be useful for HD map change detection. However, achieving robust and high-precision alignment between images and HD maps is challenging in varying environmental conditions. In addition, evaluating alignment in the absence of ground truth data is inconvenient. This article proposes a trajectory interpolation-based method for reconstructing lane markings from images to detect road changes when compared with the HD grid map. Images and HD maps are registered using a distance transform and cross-entropy-based optimization. The article also proposes a metric based on intersection over union for evaluating alignment accuracy. Experiments were conducted using a third-party-collected campus dataset and the Argoverse2 open-source dataset to demonstrate the effectiveness of the proposed methods in detecting road changes and achieving high-precision image and HD map fusion.

Proxy-CPM: A collaborative high definition map update for autonomous and connected vehicles

Proxy-CPM: A collaborative high definition map update for autonomous and connected vehicles

Proxy-CPM: a collaborative high-definition map update for autonomous and connected vehicles

Proxy-CPM: a collaborative high-definition map update for autonomous and connected vehicles

DRL-MURA: A Joint Optimization of High-Definition Map Updating and Wireless Resource Allocation in Vehicular Edge Computing Networks

DRL-MURA: A Joint Optimization of High-Definition Map Updating and Wireless Resource Allocation in Vehicular Edge Computing Networks

도로 상태 관리를 위한 정밀도로지도 기반의 도로노면지도 개발에 관한 연구

도로 상태 관리를 위한 정밀도로지도 기반의 도로노면지도 개발에 관한 연구

Multi-Session High-Definition Map-Monitoring System for Map Update

This research paper employed a multi-session framework to present an innovative approach to map monitoring within the domain of high-definition (HD) maps. The proposed methodology uses a machine learning algorithm to derive a confidence level for the detection of specific map elements in each frame and tracks the position of the element in subsequent frames. This creates a virtual belief system, which indicates the existence of the element on the HD map. To confirm the existence of the element and ensure the credibility of the map data, a reconstruction and matching technique was implemented. The notion of an expected observation area is also introduced by strategically limiting the vehicle’s observation range, thereby bolstering the detection confidence of the observed map elements. Furthermore, we leveraged data from multiple vehicles to determine the necessity for updates within specific areas, ensuring the accuracy and dependability of the map information. The validity and practicality of our approach were substantiated by real experimental data, and the monitoring accuracy exceeded 90%.

Monovision End-to-End Dual-Lane Overtaking Network without Map Assistance

Overtaking on a dual-lane road with the presence of oncoming vehicles poses a considerable challenge in the field of autonomous driving. With the assistance of high-definition maps, autonomous vehicles can plan a relatively safe trajectory for executing overtaking maneuvers. However, the creation of high-definition maps requires extensive preparation, and in rural areas where dual two-lane roads are common, there is little pre-mapping to provide high-definition maps. This paper proposes an end-to-end model called OG-Net (Overtaking Guide Net), which accomplishes overtaking tasks without map generation or communication with other vehicles. OG-Net initially evaluates the likelihood of a successful overtaking maneuver before executing the necessary actions. It incorporates the derived probability value with a set of simple parameters and utilizes a Gaussian differential controller to determine the subsequent vehicle movements. The Gaussian differential controller effectively adapts a fixed geometric curve to various driving scenarios. Unlike conventional autonomous driving models, this approach employs uncomplicated parameters rather than RNN-series networks to integrate contextual information for overtaking guidance. Furthermore, this research curated a new end-to-end overtaking dataset, CarlaLanePass, comprising first-view image sequences, overtaking success rates, and real-time vehicle status during the overtaking process. Extensive experiments conducted on diverse road scenes using the Carla platform support the validity of our model in achieving successful overtaking maneuvers.

Online Quantitative Analysis of Perception Uncertainty Based on High-Definition Map.

Environmental perception plays a fundamental role in decision-making and is crucial for ensuring the safety of autonomous driving. A pressing challenge is the online evaluation of perception uncertainty, a crucial step towards ensuring the safety and the industrialization of autonomous driving. High-definition maps offer precise information about static elements on the road, along with their topological relationships. As a result, the map can provide valuable prior information for assessing the uncertainty associated with static elements. In this paper, a method for evaluating perception uncertainty online, encompassing both static and dynamic elements, is introduced based on the high-definition map. The proposed method is as follows: Firstly, the uncertainty of static elements in perception, including the uncertainty of their existence and spatial information, was assessed based on the spatial and topological features of the static environmental elements; secondly, an online assessment model for the uncertainty of dynamic elements in perception was constructed. The online evaluation of the static element uncertainty was utilized to infer the dynamic element uncertainty, and then a model for recognizing the driving scenario and weather conditions was constructed to identify the triggering factors of uncertainty in real-time perception during autonomous driving operations, which can further optimize the online assessment model for perception uncertainty. The verification results on the nuScenes dataset show that our uncertainty assessment method based on a high-definition map effectively evaluates the real-time perception results' performance.

LINE AND POLYGON TOPOLOGY IN OPENDRIVE MODELLING

Abstract. The paper discusses the importance of efficient methods in the automotive industry for the development of self-driving vehicles and advanced vehicle assistants, focusing on the use of high-definition (HD) maps. The integration of computer simulation and HD maps in the OpenDRIVE format is emphasized. A paradigm shift in map topology is highlighted, requiring a new map creation and usage approach. The article presents a methodology that addresses both geometric and topological aspects in creating accurate HD map models. The method focuses on the connection of linear and arc road elements and eliminating continuity/connectivity errors in lane descriptions. Real-world tests validate the implemented methodology and demonstrate the successful generation of topologically correct HD map models. The results show the potential of these models for various automotive applications, particularly in the development and testing of self-driving vehicles and advanced vehicle assistants. The methodology contributes to the advancement of HD map creation, providing valuable insights for researchers and map makers in the automotive industry.

AN APPROACH OF HIGH DEFINITION MAP INFORMATION INTERACTION

Abstract. High definition (HD) maps play a very important role in the realization of autonomous driving technology. It assists self-driving vehicles to efficiently and safely complete a series of driving decisions and route planning by virtue of having most of the accurate and reliable prior information in the road environment. With the continuous change of technology, there are higher requirements for the accuracy, richness and freshness of the information stored in the HD map, so as to assist the practical application of automatic driving technology. However, current research related to HD maps mainly focuses on static information in the road environment. Since there is a large amount of complex, variable and uncertain dynamic information in the road environment, it can be used as prior knowledge for self-driving to make better decisions. Therefore, the research focus of this paper is on the dynamic information. We propose to use HD map as an information system - high definition map information system (HDMIS) - to assist autonomous driving vehicles. We design the specific content of dynamic information in the HDMIS, and develop an information interaction approach based on the vehicle end of the self-driving car and the HDMIS cloud as the interactive subject of dynamic information. In the experiment, we design and build three types of specific traffic scenarios on the simulation platform, and verify the effectiveness of the interaction approach by using the database to perform information interaction between different ports. The results show that our interaction approach can meet the storage and release of dynamic information by HDMIS to a certain extent, and can provide a large amount of dynamic information for autonomous vehicles to help them complete subsequent driving decisions and planning.

MFSCNN: APPENDING A MASKED BRANCH TO FAST-SCNN TO IMPROVE ROAD MARKING EXTRACTION ON SPARSE MLS POINT CLOUD-DERIVED IMAGES

Abstract. With the rise of self-driving cars, an increasing number of vehicles are equipped with low-cost light detection and ranging (LiDAR) sensors that could potentially serve as a massive mobile mapping resource, particularly for jobs that require multiple and frequent scanning, such as maintaining dynamic high-definition maps or digital twins. However, low-cost LiDAR sensors produce sparser point clouds during scanning which can make deep learning techniques for the automatic retrieval of features difficult like extracting road markings. In this work, we aim to improve the performance of a convolutional neural network (CNN) model for road marking extraction from sparse mobile LiDAR scanning (MLS) point cloud-derived images. We propose the modification of the Fast-SCNN model structure by adding a 2D convolution branch with masking in the feature fusion step: MFSCNN. To retain speed we only use MFSCNN to boost model training and still utilize Fast-SCNN for inference. Our results indicate potential, with a 4.6% increase in mean f1-score and an 8% decrease in uncertainty for the road marking class after multiple trials. Additionally, this research aims to support and increase research interest in lower-cost LiDARs for mobile mapping.

ESTABLISHMENT OF HD MAPS VERIFICATION AND VALIDATION PROCEDURE WITH OPENDRIVE AND AUTOWARE (LANELET2) FORMATS

Abstract. Mobile mapping technologies, for example multi-sensor integration and multi-platform mapping technology, have developed and improved over recent decades, various applications such as conventional mapping scenarios, rapid disaster response, smart city, and autonomous vehicle application arise synchronously. Especially, autonomous driving vehicles have made enormous progress. High-definition (HD) maps are key for autonomous driving because of their high accuracy and rich information of road scenes. However, how to make sure that HD maps are suitable for autonomous vehicle requirement is an important topic. The HD maps guidelines and standards in Taiwan are released since 2018 and mainly focus on point cloud and shape file format. In this paper, a procedure for the verification and validation of HD maps for OpenDRIVE and Autoware (Lanelet2) is proposed. It discusses about the verification strategies, suggestion review item, recommendation tools, and process. As shown by our preliminary results, the proposed process can conform not only in closed area but also public road. These issues can help reducing HD maps production costs. When the foundation of HD maps accomplishes, the autonomous driving techniques can naturally complement. The vision of full automation vehicle will come true rapidly in the future.

VANISHING POINT AIDED LANE DETECTION USING A MULTI-SENSOR SYSTEM

Abstract. Lane Detection is a critical component of an autonomous driving system that can be integrated alongside with High-definition (HD) map to improve accuracy and reliability of the system. Typically, lane detection is achieved using computer vision algorithms such as edge detection and Hough transform, deep learning-based algorithms, or motion-based algorithms to detect and track the lanes on the road. However, these approaches can contain incorrectly detected line segments with outliers. To address these issues, we proposed a vanishing point aided lane detection method that utilizes both camera and LiDAR sensors, and then employs a RANSAC-based post-processing method to remove potential outliers to improve the accuracy of the detected lanes. We evaluated this method on four datasets provided from the KITTI Benchmark Suite and achieved a total precision of 87%.

VEHICLE OCCLUSION REMOVAL FROM SINGLE AERIAL IMAGES USING GENERATIVE ADVERSARIAL NETWORKS

Abstract. Removing occluding objects such as vehicles from drivable areas allows precise extraction of road boundaries and related semantic objects such as lane-markings, which is crucial for several applications such as generating high-definition maps for autonomous driving. Conventionally, multiple images of the same area taken at different times or from various perspectives are used to remove occlusions and to reconstruct the occluded areas. Nevertheless, these approaches require large amounts of data, which are not always available. Furthermore, they do not work for static occlusions caused by, among others, parked vehicles. In this paper, we address occlusion removal based on single aerial images using generative adversarial networks (GANs), which are able to deal with the mentioned challenges. To this end, we adapt several state-of-the-art GAN-based image inpainting algorithms to reconstruct the missing information. Results indicate that the StructureFlow algorithm outperforms the competitors and the restorations obtained are robust, with high visual fidelity in real-world applications. Furthermore, due to the lack of annotated aerial vehicle removal datasets, we generate a new dataset for training and validating the algorithms, the Aerial Vehicle Occlusion Removal (AVOR) dataset. To the best of our knowledge, our work is the first to address vehicle removal using deep learning algorithms to enhance maps.

Open-Pit Map: An HD Map Data Model for Open-Pit Mines

Open-pit mines are becoming more intelligent and unmanned. One of the core technologies for autonomous driving is high-definition map technology, while refined mine maps are also needed to support scenarios such as vehicle–shovel collaboration, loading and unloading, production scheduling, and traffic safety control. To better serve the construction of intelligent mines and address the autonomous driving needs in open-pit mining scenarios, a universal high-definition map data model oriented toward data production and exchange, Open-Pit Map (OP map for short), has been designed. Based on various working scenarios of open-pit mines, eight practical basic elements, eleven location elements, and eight traffic elements in OP map were proposed. A map visualization tool, MapViewer, was developed based on OP map, and the effectiveness and practicality of the data model in the practical application of an intelligent shipping system in a Chinese mine were validated.