Wide deployment of global positioning system (GPS) sensors has generated a large amount of data with numerous applications in transportation research. Due to the observation error, a map matching (MM) process is commonly performed to infer a path on a road network from a noisy GPS trajectory. The increasing data volume calls for the design of efficient and scalable MM algorithms. This article presents fast map matching (FMM), an algorithm integrating hidden Markov model with precomputation, and provides an open-source implementation. An upper bounded origin-destination table is precomputed to store all pairs of shortest paths within a certain length in the road network. As a benefit, repeated routing queries known as the bottleneck of MM are replaced with hash table search. Additionally, several degenerate cases and a problem of reverse movement are identified and addressed in FMM. Experiments on a large collection of real-world taxi trip trajectories demonstrate that FMM has achieved a considerable single-processor MM speed of 25,000–45,000 points/second varying with the output mode. Investigation on the running time of different steps in FMM reveals that after precomputation is employed, the new bottleneck is located in candidate search, and more specifically, the projection of a GPS point to the polyline of a road edge. Reverse movement in the result is also effectively reduced by applying a penalty.

Next point-of-interest (POI) recommendation has been applied by many internet companies to enhance the user travel experience. Recent research advocates deep-learning methods to model long-term check-in sequences and mine mobility patterns of people to improve recommendation performance. Existing approaches model general user preferences based on historical check-ins and can be termed as preference pattern models. The preference pattern is different from the intention pattern, in that it does not emphasize the user mobility pattern of revisiting POIs, which is a common behavior and kind of intention for users. An effective module is needed to predict when and where users will repeat visits. In this paper, we propose a Spatio-Temporal Intention Learning Self-Attention Network (STILSAN) for next POI recommendation. STILSAN employs a preference-intention module to capture the user’s long-term preference and recognizes the user’s intention to revisit some specific POIs at a specific time. Meanwhile, we design a spatial encoder module as a pretrained model for learning POI spatial feature by simulating the spatial clustering phenomenon and the spatial proximity of the POIs. Experiments are conducted on two real-world check-in datasets. The experimental results demonstrate that all the proposed modules can effectively improve recommendation accuracy and STILSAN yields outstanding improvements over the state-of-the-art models.

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.

Large collections of trajectories provide rich insight into movement patterns of the tracked objects. By map matching trajectories to a road network as sequences of road edge IDs, contiguous sequential patterns can be extracted as a certain number of objects traversing a specific path, which provides valuable information in travel demand modeling and transportation planning. Mining and visualization of such patterns still face challenges in efficiency, scalability, and visual cluttering of patterns. To address these challenges, this article firstly proposes a Bidirectional Pruning based Closed Contiguous Sequential pattern Mining (BP-CCSM) algorithm. By employing tree structures to create partitions of input sequences and candidate patterns, closeness can be checked efficiently by comparing nodes in a tree. Secondly, a system called Sequential Pattern Explorer for Trajectories (SPET) is built for spatial and temporal exploration of the mined patterns. Two types of maps are designed where a conventional traffic map gives an overview of the movement patterns and a dynamic offset map presents detailed information according to user-specified filters. Extensive experiments are performed in this article. BP-CCSM is compared with three other state-of-the-art algorithms on two datasets: a small public dataset containing clickstreams from an e-commerce and a large global positioning system dataset with more than 600,000 taxi trip trajectories. The results show that BP-CCSM considerably outperforms three other algorithms in terms of running time and memory consumption. Besides, SPET provides an efficient and convenient way to inspect spatial and temporal variations in closed contiguous sequential patterns from a large number of trajectories.

Taxi demand forecasting plays an important role in ride-hailing services. Accurate taxi demand forecasting can assist taxi companies in pre-allocating taxis, improving vehicle utilization, reducing waiting time, and alleviating traffic congestion. It is a challenging task due to the highly non-linear and complicated spatial-temporal patterns of the taxi data. Most of the existing taxi demand forecasting methods lack the ability to capture the dynamic spatial-temporal dependencies among regions. They either fail to consider the limitations of Graph Neural Networks or do not efficiently capture the long-term temporal dependencies. In this paper, we propose a Spatial-Temporal Diffusion Convolutional Network (ST-DCN) for taxi demand forecasting. The dynamic spatial dependencies are efficiently captured through a two-phase graph diffusion convolutional network where the attention mechanism is introduced. Moreover, a novel temporal convolution module is designed to learn various ranges of temporal dependencies, including recent, daily, and weekly periods. Inside the module, convolution layers are stacked to handle very long sequences. Experimental results on two large-scale real-world taxi datasets from New York City (NYC) and Chengdu demonstrate that our method significantly outperforms seven state-of-the-art baseline methods.

Geocoding is an essential procedure in geographical information retrieval to associate place names with coordinates. Due to the inherent ambiguity of place names in natural language and the scarcity of place names in textual data, it is widely recognized that geocoding is challenging. Recent advances in deep learning have promoted the use of the neural network to improve the performance of geocoding. However, most of the existing approaches consider only the local context, e.g., neighboring words in a sentence, as opposed to the global context, e.g., the topic of the document. Lack of global information may have a severe impact on the robustness of the model. To fill the research gap, this paper proposes a novel global context embedding approach to generate linguistic and geospatial features through topic embedding and location embedding, respectively. A deep neural network called LGGeoCoder, which integrates local and global features, is developed to solve the geocoding as a classification problem. The experiments on a Wikipedia place name dataset demonstrate that LGGeoCoder achieves competitive performance compared with state-of-the-art models. Furthermore, the effect of introducing global linguistic and geospatial features in geocoding to alleviate the ambiguity and scarcity problem is discussed.