This page has only limited features, please log in for full access.
In the domain of air traffic control (ATC) systems, efforts to train a practical automatic speech recognition (ASR) model always faces the problem of small training samples since the collection and annotation of speech samples are expert- and domain-dependent task. In this work, a novel training approach based on pretraining and transfer learning is proposed to address this issue, and an improved end-to-end deep learning model is developed to address the specific challenges of ASR in the ATC domain. An unsupervised pretraining strategy is first proposed to learn speech representations from unlabeled samples for a certain dataset. Specifically, a masking strategy is applied to improve the diversity of the sample without losing their general patterns. Subsequently, transfer learning is applied to fine-tune a pretrained or other optimized baseline models to finally achieves the supervised ASR task. By virtue of the common terminology used in the ATC domain, the transfer learning task can be regarded as a sub-domain adaption task, in which the transferred model is optimized using a joint corpus consisting of baseline samples and new transcribed samples from the target dataset. This joint corpus construction strategy enriches the size and diversity of the training samples, which is important for addressing the issue of the small transcribed corpus. In addition, speed perturbation is applied to augment the new transcribed samples to further improve the quality of the speech corpus. Three real ATC datasets are used to validate the proposed ASR model and training strategies. The experimental results demonstrate that the ASR performance is significantly improved on all three datasets, with an absolute character error rate only one-third of that achieved through the supervised training. The applicability of the proposed strategies to other ASR approaches is also validated.
Yi Lin; Qin Li; Bo Yang; Zhen Yan; Huachun Tan; Zhengmao Chen. Improving speech recognition models with small samples for air traffic control systems. Neurocomputing 2021, 445, 287 -297.
AMA StyleYi Lin, Qin Li, Bo Yang, Zhen Yan, Huachun Tan, Zhengmao Chen. Improving speech recognition models with small samples for air traffic control systems. Neurocomputing. 2021; 445 ():287-297.
Chicago/Turabian StyleYi Lin; Qin Li; Bo Yang; Zhen Yan; Huachun Tan; Zhengmao Chen. 2021. "Improving speech recognition models with small samples for air traffic control systems." Neurocomputing 445, no. : 287-297.
Highway system is experiencing increasing traffic congestion with fast-growing number of vehicles in metropolitan areas. Implementing traffic management strategies such as utilizing the hard shoulder as an extra lane could increase highway capacity without extra construction work. This paper presents a method of determining an optimal traffic condition and speed limit of opening hard shoulder. Firstly, the traffic states are clustered using K-Means, mean shift, agglomerative and spectral clustering methods, and the optimal clustering algorithm is selected using indexes including the silhouette score, Davies-Bouldin Index and Caliski-Harabaz Score. The results suggested that the clustering effect of using K-Means method with three categories is optimal. Then, cellular automata model is used to simulate traffic conditions before and after the hard shoulder running strategy is applied. The parameters of the model, including the probabilities of random deceleration, slow start and lane change, are calibrated using real traffic data. Four indicators including the traffic volume, the average speed, the variance of speed, and the travel time of emergency rescue vehicles during traffic accident obtained using the cellular automata model are used to evaluate various hard shoulder running strategies. By using factor analysis and TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) methods, the optimal traffic condition and speed limit of opening hard shoulder could be determined. This method could be applied to highway segments of various number of lanes and different speed limits to optimize the hard shoulder running strategy for highway management
Fan Yang; Fan Wang; Fan Ding; Huachun Tan; Bin Ran. Identify Optimal Traffic Condition and Speed Limit for Hard Shoulder Running Strategy. Sustainability 2021, 13, 1822 .
AMA StyleFan Yang, Fan Wang, Fan Ding, Huachun Tan, Bin Ran. Identify Optimal Traffic Condition and Speed Limit for Hard Shoulder Running Strategy. Sustainability. 2021; 13 (4):1822.
Chicago/Turabian StyleFan Yang; Fan Wang; Fan Ding; Huachun Tan; Bin Ran. 2021. "Identify Optimal Traffic Condition and Speed Limit for Hard Shoulder Running Strategy." Sustainability 13, no. 4: 1822.
Car-following models have been widely applied and made remarkable achievements in traffic engineering. However, the traffic micro-simulation accuracy of car-following models in a platoon level, especially during traffic oscillations, still needs to be enhanced. Rather than using traditional individual car-following models, we proposed a new trajectory generation approach to generate platoon level trajectories given the first leading vehicle's trajectory. In this article, we discussed the temporal and spatial error propagation issue for the traditional approach by a car following block diagram representation. Based on the analysis, we pointed out that error comes from the training method and the model structure. In order to fix that, we adopt two improvements on the basis of the traditional LSTM-based car-following model. We utilized a scheduled sampling technique during the training process to solve the error propagation in the temporal dimension. Furthermore, we developed a unidirectional interconnected LSTM model structure to extract trajectories features from the perspective of the platoon. As indicated by the systematic empirical experiments, the proposed novel structure could efficiently reduce the temporal-spatial error propagation. Compared with the traditional LSTM-based car-following model, the proposed model has almost 40% less error. The findings will benefit the design and analysis of micro-simulation for platoon-level car-following models.
Yangxin Lin; Ping Wang; Yang Zhou; Fan Ding; Chen Wang; Huachun Tan. Platoon Trajectories Generation: A Unidirectional Interconnected LSTM-Based Car-Following Model. IEEE Transactions on Intelligent Transportation Systems 2020, PP, 1 -11.
AMA StyleYangxin Lin, Ping Wang, Yang Zhou, Fan Ding, Chen Wang, Huachun Tan. Platoon Trajectories Generation: A Unidirectional Interconnected LSTM-Based Car-Following Model. IEEE Transactions on Intelligent Transportation Systems. 2020; PP (99):1-11.
Chicago/Turabian StyleYangxin Lin; Ping Wang; Yang Zhou; Fan Ding; Chen Wang; Huachun Tan. 2020. "Platoon Trajectories Generation: A Unidirectional Interconnected LSTM-Based Car-Following Model." IEEE Transactions on Intelligent Transportation Systems PP, no. 99: 1-11.
With the precedence of connected automated vehicles (CAVs), car-following control technology is a promising way to enhance traffic safety. Although a variety of research has been conducted to analyze the safety enhancement by CAV technology, the parametric impact on CAV technology has not been systematically explored. Hence, this paper analyzes the parametric impacts on surrogate safety measures (SSMs) for a mixed vehicular platoon via a two-level analysis structure. To construct the active safety evaluation framework, numerical simulations were constructed which can generate trajectories for different kind of vehicles while considering communication and vehicle dynamics characteristics. Based on the trajectories, we analyzed parametric impacts upon active safety on two different levels. On the microscopic level, parameters including controller dynamic characteristics and equilibrium time headway of car-following policies were analyzed, which aimed to capture local and aggregated driving behavior’s impact on the vehicle. On the macroscopic level, parameters incorporating market penetration rate (MPR), vehicle topology, and vehicle-to-vehicle environment were extensively investigated to evaluate their impacts on aggregated platoon level safety caused by inter-drivers’ behavioral differences. As indicated by simulation results, an automated vehicle (AV) suffering from degradation is a potentially unsafe component in platoon, due to the loss of a feedforward control mechanism. Hence, the introduction of connected automated vehicles (CAVs) only start showing benefits to platoon safety from about 20% CAV MPR in this study. Furthermore, the analysis on vehicle platoon topology suggests that arranging all CAVs at the front of a mixed platoon assists in enhancing platoon SSM performances.
Fan Ding; Jiwan Jiang; Yang Zhou; Ran Yi; Huachun Tan. Unravelling the Impacts of Parameters on Surrogate Safety Measures for a Mixed Platoon. Sustainability 2020, 12, 9955 .
AMA StyleFan Ding, Jiwan Jiang, Yang Zhou, Ran Yi, Huachun Tan. Unravelling the Impacts of Parameters on Surrogate Safety Measures for a Mixed Platoon. Sustainability. 2020; 12 (23):9955.
Chicago/Turabian StyleFan Ding; Jiwan Jiang; Yang Zhou; Ran Yi; Huachun Tan. 2020. "Unravelling the Impacts of Parameters on Surrogate Safety Measures for a Mixed Platoon." Sustainability 12, no. 23: 9955.
Understanding the underlying patterns of the urban mobility dynamics is essential for both the traffic state estimation and management of urban facilities and services. Due to the coupling relationship of generative factors in spatial-temporal domain, it is challenging to model the citywide traffic dynamics under a structural pattern of critical features such as hours of days, days of weeks and weather conditions. To address this challenge, this article develops a disentangled representation learning framework to learn an interpretable factorized representation of the independent data generative factors. In order to make full use of the knowledge on generative factors, this article proposes spatial-temporal generative adversarial network (ST-GAN) to assign the generative factors of traffic flow to the feature vector in latent space and reconstructs the high-dimensional citywide traffic flow from the given factors. With the help of the disentangled representations, the decomposed feature vector in latent space discloses the relationship between underlying patterns and citywide traffic dynamics. Several comprehensively experiments show that ST-GAN not only effectively improves the prediction accuracy but also promisingly characterize structural properties of the traffic evolution process.
Hailong Zhang; Yuankai Wu; Huachun Tan; Hanxuan Dong; Fan Ding; Bin Ran. Understanding and Modeling Urban Mobility Dynamics via Disentangled Representation Learning. IEEE Transactions on Intelligent Transportation Systems 2020, PP, 1 -11.
AMA StyleHailong Zhang, Yuankai Wu, Huachun Tan, Hanxuan Dong, Fan Ding, Bin Ran. Understanding and Modeling Urban Mobility Dynamics via Disentangled Representation Learning. IEEE Transactions on Intelligent Transportation Systems. 2020; PP (99):1-11.
Chicago/Turabian StyleHailong Zhang; Yuankai Wu; Huachun Tan; Hanxuan Dong; Fan Ding; Bin Ran. 2020. "Understanding and Modeling Urban Mobility Dynamics via Disentangled Representation Learning." IEEE Transactions on Intelligent Transportation Systems PP, no. 99: 1-11.
Trip generation modeling is essential in transportation planning activities. Previous modeling methods that depend on traditional data collection methods are inefficient and expensive. This paper proposed a novel data-driven trip generation modeling method for urban residents and non-local travelers utilizing location-based social network (LBSN) data and cellular phone data and conducted a case study in Nanjing, China. First, the point of interest (POI) data of the LBSN were classified into various categories by the service type, then, four features of each category including the number of users, number of POIs, number of check-ins, and number of photos were aggregated by traffic analysis zones to be used as explanatory variables for the trip generation models. We used a random tree regression method to select the most important features as the model inputs, and the trip models were established based on the ordinary least square model. Then, an exploratory approach was used to test the performance of each combination of the variables with various test methods to identify the best model for residents’ and travelers’ trip generation functions. The results suggest land use compositions have significant impact on trip generations, and the trip generation patterns are different between urban residents and non-local travelers.
Fan Yang; Linchao Li; Fan Ding; Huachun Tan; Bin Ran. A Data-Driven Approach to Trip Generation Modeling for Urban Residents and Non-local Travelers. Sustainability 2020, 12, 7688 .
AMA StyleFan Yang, Linchao Li, Fan Ding, Huachun Tan, Bin Ran. A Data-Driven Approach to Trip Generation Modeling for Urban Residents and Non-local Travelers. Sustainability. 2020; 12 (18):7688.
Chicago/Turabian StyleFan Yang; Linchao Li; Fan Ding; Huachun Tan; Bin Ran. 2020. "A Data-Driven Approach to Trip Generation Modeling for Urban Residents and Non-local Travelers." Sustainability 12, no. 18: 7688.
A driving cycle is important to accomplish an accurate depiction of a vehicle’s driving characteristics as the traction motor’s flexible response to stop and start commands. In this paper, the driving cycle construction of an urban hybrid electric bus (HEB) in Zhengzhou, China is developed in which a measurement system integrating global positioning and inertial navigation function is used to acquire driving data. The collected data are then divided into acceleration, deceleration, uniform, and stop fragments. Meanwhile, the velocity fragments are classified into seven state clusters according to their average velocities. A transfer matrix applied to reveal the transfer relationship of velocity clusters can be obtained with statistical analysis. In the third stage, a three-part construction method of driving cycle is designed. Firstly, according to the theory of Markov chain, all the alternative parts that satisfy the construction’s precondition are selected based on the transfer matrix and Monte Carlo method. The Zhengzhou urban driving cycle (ZZUDC) could be determined by comparing the performance measure (PM) values subsequently. Eventually, the method and the cycle are validated by the high correlation coefficient (0.9972) with original data of ZZUDC than that of the other driving cycle (0.9746) constructed with traditional micro-trip and as well by comparing several statistical characteristics of ZZUDC and seven international cycles. Particularly, with around 20.5 L/100 km fuel and approximately 12.8 kwh/100 km electricity consumption, there is a narrow gap between the energy consumption of ZZUDC and WVUCITY, and their characteristics are similar.
Jiankun Peng; Jiwan Jiang; Fan Ding; Huachun Tan. Development of Driving Cycle Construction for Hybrid Electric Bus: A Case Study in Zhengzhou, China. Sustainability 2020, 12, 7188 .
AMA StyleJiankun Peng, Jiwan Jiang, Fan Ding, Huachun Tan. Development of Driving Cycle Construction for Hybrid Electric Bus: A Case Study in Zhengzhou, China. Sustainability. 2020; 12 (17):7188.
Chicago/Turabian StyleJiankun Peng; Jiwan Jiang; Fan Ding; Huachun Tan. 2020. "Development of Driving Cycle Construction for Hybrid Electric Bus: A Case Study in Zhengzhou, China." Sustainability 12, no. 17: 7188.
Modern automotive systems have equipped with a highly increasing number of computer vision hardware/softwares, which are considered to be beneficial for eco-driving. This work combines computer vision and deep reinforcement learning to improve the fuel economy of hybrid electric vehicles. The convolutional neural networks-based object detection method is utilized to extract available visual information from on-board cameras. The visual information is used as a state input for a continuous deep reinforcement learning model to output energy management strategies. In order to evaluate the proposed method, we construct 100 kilometers real city and highway driving cycles, in which visual information is incorporated. The results show that the deep reinforcement learning based system with visual information consumes 4.3% to 8.8% less fuel compared with the one without visual information, and achieves 96.5% fuel economy of the dynamic programming.
Yong Wang; Huachun Tan; Yuankai Wu; Jiankun Peng. Hybrid Electric Vehicle Energy Management With Computer Vision and Deep Reinforcement Learning. IEEE Transactions on Industrial Informatics 2020, 17, 3857 -3868.
AMA StyleYong Wang, Huachun Tan, Yuankai Wu, Jiankun Peng. Hybrid Electric Vehicle Energy Management With Computer Vision and Deep Reinforcement Learning. IEEE Transactions on Industrial Informatics. 2020; 17 (6):3857-3868.
Chicago/Turabian StyleYong Wang; Huachun Tan; Yuankai Wu; Jiankun Peng. 2020. "Hybrid Electric Vehicle Energy Management With Computer Vision and Deep Reinforcement Learning." IEEE Transactions on Industrial Informatics 17, no. 6: 3857-3868.
This paper presents a personalized stochastic optimal adaptive cruise control (ACC) algorithm for automated vehicles (AVs) incorporating human drivers’ risk-sensitivity under system and measurement uncertainties. The proposed controller is designed as a linear exponential-of-quadratic Gaussian (LEQG) problem, which utilizes the stochastic optimal control mechanism to feedback the deviation from the design car-following target. With the risk-sensitive parameter embedded in LEQG, the proposed method has the capability to characterize risk preference heterogeneity of each AV against uncertainties according to each human drivers’ preference. Further, the established control theory can achieve both expensive control mode and non-expensive control mode via changing the weighting matrix of the cost function in LEQG to reveal different treatments on input. Simulation tests validate the proposed approach can characterize different driving behaviors and its effectiveness in terms of reducing the deviation from equilibrium state. The ability to produce different trajectories and generate smooth control of the proposed algorithm is also verified.
Jiwan Jiang; Fan Ding; Yang Zhou; Jiaming Wu; Huachun Tan. A Personalized Human Drivers’ Risk Sensitive Characteristics Depicting Stochastic Optimal Control Algorithm for Adaptive Cruise Control. IEEE Access 2020, 8, 145056 -145066.
AMA StyleJiwan Jiang, Fan Ding, Yang Zhou, Jiaming Wu, Huachun Tan. A Personalized Human Drivers’ Risk Sensitive Characteristics Depicting Stochastic Optimal Control Algorithm for Adaptive Cruise Control. IEEE Access. 2020; 8 ():145056-145066.
Chicago/Turabian StyleJiwan Jiang; Fan Ding; Yang Zhou; Jiaming Wu; Huachun Tan. 2020. "A Personalized Human Drivers’ Risk Sensitive Characteristics Depicting Stochastic Optimal Control Algorithm for Adaptive Cruise Control." IEEE Access 8, no. : 145056-145066.
Variable speed limit (VSL) control is a flexible way to improve traffic conditions, increase safety, and reduce emissions. There is an emerging trend of using reinforcement learning methods for VSL control. Currently, deep learning is enabling reinforcement learning to develop autonomous control agents for problems that were previously intractable. In this paper, a more effective deep reinforcement learning (DRL) model is developed for differential variable speed limit (DVSL) control, in which dynamic and distinct speed limits among lanes can be imposed. The proposed DRL model uses a novel actor-critic architecture to learn a large number of discrete speed limits in a continuous action space. Different reward signals, such as total travel time, bottleneck speed, emergency braking, and vehicular emissions are used to train the DVSL controller, and a comparison between these reward signals is conducted. The proposed DRL-based DVSL controllers are tested on a freeway with a simulated recurrent bottleneck. The simulation results show that the DRL based DVSL control strategy is able to improve the safety, efficiency and environment-friendliness of the freeway. In order to verify whether the controller generalizes to real world implementation, we also evaluate the generalization of the controllers on environments with different driving behavior attributes. and the robustness of the DRL agent is observed from the results.
Yuankai Wu; Huachun Tan; Lingqiao Qin; Bin Ran. Differential variable speed limits control for freeway recurrent bottlenecks via deep actor-critic algorithm. Transportation Research Part C: Emerging Technologies 2020, 117, 102649 .
AMA StyleYuankai Wu, Huachun Tan, Lingqiao Qin, Bin Ran. Differential variable speed limits control for freeway recurrent bottlenecks via deep actor-critic algorithm. Transportation Research Part C: Emerging Technologies. 2020; 117 ():102649.
Chicago/Turabian StyleYuankai Wu; Huachun Tan; Lingqiao Qin; Bin Ran. 2020. "Differential variable speed limits control for freeway recurrent bottlenecks via deep actor-critic algorithm." Transportation Research Part C: Emerging Technologies 117, no. : 102649.
Inferring travel modes of travelers in the city is important to transportation planning and infrastructure design. Based on the distribution of travel modes, transportation engineers could provide some proper strategies to reduce traffic congestion and air pollution. With advanced sensing techniques, it is possible to collect high-resolution GPS trajectory data of travelers and we can infer travel modes using some popular machine learning methods. One of the difficult tasks facing the application of machine learning especially deep learning in travel mode detection is the lack of large, labeled dataset, because to label the trajectory data is expensive and time-consuming. Moreover, samples of different travel modes are always unbalanced. Accordingly, in this paper, we take advantage of the generative model and the Convolutional Neural Networks (CNN) to develop a hybrid travel modes detection model using less labeled trajectory data. Our key contribution is the utilization of a generative adversarial network (GAN) to artificially create some training samples in such a way that it not only increases the required sample size but balances the dataset to improve the accuracy of the detection model. Furthermore, CNN is applied to extract deep features of trajectory data, and then to classify the travel modes. The results show that the highest accuracy (86.70%) can be achieved by the proposed model. In particular, the proposed method can improve the detection accuracy of bus and driving modes because it can solve the small sample size problem. Moreover, the large sample size can provide an opportunity to develop some advanced deep learning models in future studies.
Linchao Li; Jiasong Zhu; Hailong Zhang; Huachun Tan; Bowen Du; Bin Ran. Coupled application of generative adversarial networks and conventional neural networks for travel mode detection using GPS data. Transportation Research Part A: Policy and Practice 2020, 136, 282 -292.
AMA StyleLinchao Li, Jiasong Zhu, Hailong Zhang, Huachun Tan, Bowen Du, Bin Ran. Coupled application of generative adversarial networks and conventional neural networks for travel mode detection using GPS data. Transportation Research Part A: Policy and Practice. 2020; 136 ():282-292.
Chicago/Turabian StyleLinchao Li; Jiasong Zhu; Hailong Zhang; Huachun Tan; Bowen Du; Bin Ran. 2020. "Coupled application of generative adversarial networks and conventional neural networks for travel mode detection using GPS data." Transportation Research Part A: Policy and Practice 136, no. : 282-292.
With the development of sensing technology, a large amount of heterogeneous traffic data can be collected. However, the raw data often contain corrupted or missing values, which need to be imputed to aid traffic condition monitoring and the assessment of the system performance. Several existing studies have reported imputation models used to impute the missing values, and most of these models aimed to capture the spatial or temporal dependencies. However, the dependencies of the heterogeneous data were ignored. To this end, we propose a multimodal deep learning model to enable heterogeneous traffic data imputation. The model involves the use of two parallel stacked autoencoders that can simultaneously consider the spatial and temporal dependencies. In addition, a latent feature fusion layer is developed to capture the dependencies of the heterogeneous traffic data. To train the proposed imputation model, a hierarchical training method is introduced. Using a real world dataset, the performance of the proposed model is evaluated and compared with that of several widely used temporal imputation models, spatial imputation models, and spatial–temporal imputation models. The experimental and evaluation results indicate that the values of the evaluation criteria of the proposed model are smaller, indicating a better performance. The results also show that the proposed model can accurately impute the continuously missing data. Furthermore, the sensitivity of the parameters used in the proposed deep multimodal deep learning model is investigated. This study clearly demonstrates the effectiveness of deep learning for heterogeneous traffic data synthesis and missing data imputation. The dependencies of the heterogeneous traffic data should be considered in future studies to improve the performance of the imputation model.
Linchao Li; Bowen Du; Yonggang Wang; Lingqiao Qin; Huachun Tan. Estimation of missing values in heterogeneous traffic data: Application of multimodal deep learning model. Knowledge-Based Systems 2020, 194, 105592 .
AMA StyleLinchao Li, Bowen Du, Yonggang Wang, Lingqiao Qin, Huachun Tan. Estimation of missing values in heterogeneous traffic data: Application of multimodal deep learning model. Knowledge-Based Systems. 2020; 194 ():105592.
Chicago/Turabian StyleLinchao Li; Bowen Du; Yonggang Wang; Lingqiao Qin; Huachun Tan. 2020. "Estimation of missing values in heterogeneous traffic data: Application of multimodal deep learning model." Knowledge-Based Systems 194, no. : 105592.
The rapid development of urban metropolises has attracted a growing number of immigrants and travelers, increasing the burden on transportation systems. Previous research on urban mobility patterns have ignored the temporal variations and heterogeneity in divergent urban trip makers due to the limited data resolution and coverage. In this paper, we analyzed cellular phone data of more than five million travelers for one month in Nanjing, China and proposed a method to extract trip origin and destination information from cellular phone signal data. We found that mobility patterns are different for urban residents, short-term travelers, and transfer travelers, and that trip length distributions can best be described by gamma and exponential distributions. In addition to the daily trip length distribution models, we utilized the agglomerative hieratical clustering method in order to group similar hourly trip patterns and further proposed within-day trip length distribution models under different times of the day and days of the week.
Fan Yang; Zhenxing Yao; Fan Ding; Huachun Tan; Bin Ran. Understanding Urban Mobility Pattern with Cellular Phone Data: A Case Study of Residents and Travelers in Nanjing. Sustainability 2019, 11, 5502 .
AMA StyleFan Yang, Zhenxing Yao, Fan Ding, Huachun Tan, Bin Ran. Understanding Urban Mobility Pattern with Cellular Phone Data: A Case Study of Residents and Travelers in Nanjing. Sustainability. 2019; 11 (19):5502.
Chicago/Turabian StyleFan Yang; Zhenxing Yao; Fan Ding; Huachun Tan; Bin Ran. 2019. "Understanding Urban Mobility Pattern with Cellular Phone Data: A Case Study of Residents and Travelers in Nanjing." Sustainability 11, no. 19: 5502.