Lane-level road networks are crucial components of high-precision maps and play a significant role in intelligent transportation systems. Extracting lane-level road networks at intersections presents considerable challenges due to the complex structures of intersections and diverse driving behaviors. A graph-learning based method is proposed for extracting lanes from high-precision trajectories at road intersections. A trajectory relation graph is designed to encode the directional, shape, and distance features of trajectories, capturing both the intrinsic and extrinsic relationships between trajectories. Subsequently, a Graph Transformer Network is developed to extract a representative subset of trajectories as lanes. To alleviate the problem of generating missing and extraneous lanes, a set-based lane extraction loss is introduced to achieve implicit pruning of redundancy through the attention mechanism. Comprehensive experimental results demonstrate that the proposed method outperforms state-of-the-art methods in three positional and topological accuracy metrics. The method achieves lane extraction with minimal omissions and redundancies and exhibits strong performance in complex scenarios such as U-turns, lane merging, and lane diverging regions.

Transitions of spatial data infrastructures (SDIs) support applications from 2D landscapes to 3D scenes. The existing methods for describing, managing, and providing services for 3D spatial data often lack coordination and efficiency. Moreover, the added complexity of 3D data structures necessitates novel approaches for component-level management and streaming capabilities. In response, we developed a generic conceptual model suitable for component-level management of diverse 3D spatial data in SDIs and discussed the design rationales and key considerations underlying the model. We formalized the flexible data composition and fine-grained lifecycle management in this model and specified this model at the cloud-optimized encoding level to enable efficient CRUD operations and streaming delivery of massive 3D spatial data. Our approach enabled direct streaming of the managed 3D spatial data without the need for redundant replication. We implemented, evaluated, and discussed the proposed approach in terms of service, accessibility, visualization, analysis cases, and efficiency. The results show that the proposed method is efficient in managing 3D spatial data and enables users to conduct 3D geo-analysis on the basis of specific parts of the data as needed. This work provides a scientific exploration that integrates the management and services of 3D spatial data in SDIs.

Ensuring the correctness of road network data is critical for navigation, traffic control and urban planning. Errors like missing roads and absent connections can hinder its quality. Trajectory data emerges as a cost-effective source to uncover such errors. Existing methods often analyze the mismatches between trajectories and road networks to identify specific errors. They heavily rely on manually established rules and fail to fully leverage the diverse patterns of trajectories and the underlying road network structure. The article introduces a sequential classification approach to detect diverse road network errors. It starts with partial map matching (PMM) to associate trajectories with a road network, allowing unmatched portions. Context features are subsequently extracted by encoding patterns in the map matching (MM) outputs, raw trajectories and road network. Finally, a bidirectional recurrent neural network (BiRNN) model is trained to identify the network error category for each trajectory point. Experiments were performed on detecting errors in OpenStreetMap (OSM) road network with a real-world trajectory dataset. It demonstrates that the proposed method achieves accuracy over 96%, significantly surpassing four baselines. An ablation study confirms the necessity of considering different types of context features. This method advances error detection by effectively utilizing trajectories in identifying diverse network errors.

–Recent advances in mobile mapping systems have facilitated the collection of high-precision trajectory data in centimeter positioning accuracy. It provides the potential to infer lane-level road networks, which are essential for autonomous driving navigation. This task is challenging due to the complicated lane merging and diverging structures as well as the lane-changing patterns in trajectory data. This paper presents a lane-level road network generation method from high-precision trajectory data with lane-changing behavior analysis. Trajectories are firstly partitioned by detecting road intersections and changes in lane structure. Subsequently, in regions with consistent lane structure, a principal curve fitting algorithm is developed to extract lane centerlines. Erroneous lanes generated by lane-changing behavior are pruned based on a constructed lane intersection graph. In regions with merging and diverging lanes, a lane-group fitting algorithm is designed. This algorithm estimates lane locations by incorporating a Gaussian mixture model with lane width prior knowledge and then infers lane-level topological structures using trajectory flow information. The proposed method is evaluated on a real-world high-precision trajectory dataset. Comprehensive experiments demonstrate that it outperforms state-of-the-art methods in four metrics. Under complex scenarios, the method is capable of generating lane-level road networks with higher completeness and fewer fragments.

The popularity of ride-hailing platforms has significantly improved travel efficiency by providing convenient and personalized transportation services. Designing an effective ride-hailing service generally needs to address two tasks: order matching that assigns orders to available vehicles and proactive vehicle repositioning that deploys idle vehicles to potentially high-demand regions. Recent studies have intensively utilized deep reinforcement learning to solve the two tasks by learning an optimal dispatching strategy. However, most of them generate actions for the two tasks independently, neglecting the interactions between the two tasks and the communications among multiple drivers. To this end, this paper provides an approach based on multi-agent deep reinforcement learning where the two tasks are modeled as a unified Markov decision process, and the colossal state space and competition among drivers are addressed. Additionally, a modifiable agent-specific state representation is proposed to facilitate knowledge transferring and improve computing efficiency. We evaluate our approach on a public taxi order dataset collected in Chengdu, China, where a variable number of simulated vehicles are tested. Experimental results show that our approach outperforms seven existing baselines, reducing passenger rejection rate, driver idle time and improving total driver income.

Data management and analysis are challenging with big Earth observation (EO) data. Expanding upon the rising promises of data cubes for analysis-ready big EO data, we propose a new geospatial infrastructure layered over a data cube to facilitate big EO data management and analysis. Compared to previous work on data cubes, the proposed infrastructure, GeoCube, extends the capacity of data cubes to multi-source big vector and raster data. GeoCube is developed in terms of three major efforts: formalize cube dimensions for multi-source geospatial data, process geospatial data query along these dimensions, and organize cube data for high-performance geoprocessing. This strategy improves EO data cube management and keeps connections with the business intelligence cube, which provides supplementary information for EO data cube processing. The paper highlights the major efforts and key research contributions to online analytical processing for dimension formalization, distributed cube objects for tiles, and artificial intelligence enabled prediction of computational intensity for data cube processing. Case studies with data from Landsat, Gaofen, and OpenStreetMap demonstrate the capabilities and applicability of the proposed infrastructure.

Artificial Intelligence Machine Learning (AI/ML), in particular Deep Learning (DL), is reorienting and transforming Earth Observation (EO). A consistent data model for delivery of training data will support the FAIR data principles (findable, accessible, interoperable, reusable) and enable Web-based use of training data in a spatial data infrastructure (SDI). Existing training datasets, including open source benchmark datasets, are usually packaged into public or personal repositories and lack discoverability and accessibility. Moreover, there is no unified method to describe the training data. Here we propose a training data model for AI in EO to allow documentation, storage, and sharing of geospatial training data in a distributed infrastructure. We present design rationales, information models, and an encoding method. Several scenarios illustrate the intended uses and benefits for EO DL applications in an open Web environment. The relationship with Open Geospatial Consortium (OGC) standards is also discussed, as is the impact on an AI-ready SDI.

High performance computing is required for fast geoprocessing of geospatial big data. Using spatial domains to represent computational intensity (CIT) and domain decomposition for parallelism are prominent strategies when designing parallel geoprocessing applications. Traditional domain decomposition is limited in evaluating the computational intensity, which often results in load imbalance and poor parallel performance. From the data science perspective, machine learning from Artificial Intelligence (AI) shows promise for better CIT evaluation. This paper proposes a machine learning approach for predicting computational intensity, followed by an optimized domain decomposition, which divides the spatial domain into balanced subdivisions based on the predicted CIT to achieve better parallel performance. The approach provides a reference framework on how various machine learning methods including feature selection and model training can be used in predicting computational intensity and optimizing parallel geoprocessing against different cases. Some comparative experiments between the approach and traditional methods were performed using the two cases, DEM generation from point clouds and spatial intersection on vector data. The results not only demonstrate the advantage of the approach, but also provide hints on how traditional GIS computation can be improved by the AI machine learning.