Pub Date : 2025-09-01DOI: 10.26599/JICV.2025.9210058
Hugues Blache;Pierre-Antoine Laharotte;Nour-Eddin El Faouzi
At the dawn of the deployment of connected and automated vehicles (CAVs) on our roads, assessing the safety of new systems is crucial. Given the overwhelming number of situations to test, focusing efforts on the most relevant ones for the system is essential. Qualifying scenarios with respect to their relevance is a challenging task. The scope of relevancy must be defined, and a labeling process applicable to any scenario must be developed. However, gathering information on various scenarios to label them poses a challenge because the flagrant lacks field data. In this study, we assume that relevancy is depicted by a safety criticality level on the basis of time-to-collision. We develop a labeling process for scenarios. It learns latent connections between the words generating scenarios and takes advantage of the latent structure to associate criticality levels with any scenario. Such a prediction model enables one to cope with the lack of data by ensuring the prior qualification of any scenario regardless of the quantity of field observations. This process is applied to scenarios described at a high level of abstraction, called functional scenarios. Criticality levels might be used to guide the application of the sampling strategy to select the scenarios under consideration when testing CAVs. Compared with field observations, the results of our automated process are highly correlated, with $R^2$ values of up to 0.835 on average.
{"title":"Automatic Labeling and Qualification of Functional Scenarios on the Basis of Sparse Field Observations","authors":"Hugues Blache;Pierre-Antoine Laharotte;Nour-Eddin El Faouzi","doi":"10.26599/JICV.2025.9210058","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210058","url":null,"abstract":"At the dawn of the deployment of connected and automated vehicles (CAVs) on our roads, assessing the safety of new systems is crucial. Given the overwhelming number of situations to test, focusing efforts on the most relevant ones for the system is essential. Qualifying scenarios with respect to their relevance is a challenging task. The scope of relevancy must be defined, and a labeling process applicable to any scenario must be developed. However, gathering information on various scenarios to label them poses a challenge because the flagrant lacks field data. In this study, we assume that relevancy is depicted by a safety criticality level on the basis of time-to-collision. We develop a labeling process for scenarios. It learns latent connections between the words generating scenarios and takes advantage of the latent structure to associate criticality levels with any scenario. Such a prediction model enables one to cope with the lack of data by ensuring the prior qualification of any scenario regardless of the quantity of field observations. This process is applied to scenarios described at a high level of abstraction, called functional scenarios. Criticality levels might be used to guide the application of the sampling strategy to select the scenarios under consideration when testing CAVs. Compared with field observations, the results of our automated process are highly correlated, with <tex>$R^2$</tex> values of up to 0.835 on average.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 3","pages":"9210058-1-9210058-16"},"PeriodicalIF":7.8,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11215930","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.26599/JICV.2025.9210062
Miklós Lukovics;Barbara Nagy
Using an autonomous vehicle fleet instead of owning a car might significantly reduce the number of vehicles in cities, which may have important consequences related to land use and the urban landscape. More information is available about these possibilities; at the same time, much less is known about whether urban residents would accept them. Moreover, the majority of research addressing the preferences of urban residents presents findings on the entire population rather than on its specific sections; thus, scarce information is available about the urban landscape preferences of young people, who are highly exposed to autonomous vehicle-driven future mobility. This study aims to determine how much young city dwellers accept potential specific urban landscape changes triggered by autonomous vehicles. The research applied real-time eye-tracking tests and supplementary questionnaires to a sample of 102 participants. The tests were carried out under laboratory conditions, during which the subjects looked at before/after urban landscape pairs of images that depicted the potential urban landscape and land use changes triggered by the mass adoption of autonomous vehicles. The examination of the total fixation duration, the average fixation duration and the average number of fixations indicates that the “after” images were collectively more appealing to the subjects. An analysis of the reasons behind the eye-tracking results revealed that safety and human-centered design were identified as the most significant factors across various image pairs.
{"title":"Young Urban Dwellers' Acceptance of Autonomous Vehicle-Induced Landscape Changes: An Eye-Tracking Study","authors":"Miklós Lukovics;Barbara Nagy","doi":"10.26599/JICV.2025.9210062","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210062","url":null,"abstract":"Using an autonomous vehicle fleet instead of owning a car might significantly reduce the number of vehicles in cities, which may have important consequences related to land use and the urban landscape. More information is available about these possibilities; at the same time, much less is known about whether urban residents would accept them. Moreover, the majority of research addressing the preferences of urban residents presents findings on the entire population rather than on its specific sections; thus, scarce information is available about the urban landscape preferences of young people, who are highly exposed to autonomous vehicle-driven future mobility. This study aims to determine how much young city dwellers accept potential specific urban landscape changes triggered by autonomous vehicles. The research applied real-time eye-tracking tests and supplementary questionnaires to a sample of 102 participants. The tests were carried out under laboratory conditions, during which the subjects looked at before/after urban landscape pairs of images that depicted the potential urban landscape and land use changes triggered by the mass adoption of autonomous vehicles. The examination of the total fixation duration, the average fixation duration and the average number of fixations indicates that the “after” images were collectively more appealing to the subjects. An analysis of the reasons behind the eye-tracking results revealed that safety and human-centered design were identified as the most significant factors across various image pairs.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 3","pages":"9210062-1-9210062-13"},"PeriodicalIF":7.8,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11215928","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01DOI: 10.26599/JICV.2025.9210055
Faizan Zaman;Zhigang Xu;Adil Hussain;Anees Ullah;Khalid Zaman
This research addresses the challenging task of classifying drivers' emotions to increase their awareness of their driving behaviors. It recognizes the common issue of driver emotions, which often leads to the neglect of poor driving practices. By automatically detecting and identifying these behaviors, drivers can proactively obtain valuable insights to reduce potential accidents. This study proposes a comprehensive facial recognition model for drivers that uses a unified architecture comprising a convolutional neural network (CNN), a recurrent neural network (RNN), and a multilayer perceptron (MLP) classification model. Initially, a faster region-based convolutional neural network (R-CNN) was employed for accurate and efficient facial detection of drivers in live and recorded videos. Features are extracted from three CNN models and merged via advanced techniques to create an ensemble classification model. Moreover, the improved Faster R-CNN feature learning module is replaced with a new convolutional neural network module, VGG16, which maximizes the precision and effectiveness of facial detection in our system. Significant accuracy results of 89.2%, 97.20%, 99.01%, 93.65%, and 98.61% are shown in evaluations of our suggested facial detection and facial expression recognition (DFER) datasets, including the EMOTIC, CK+, FERPLUS, AffectNet, and custom datasets. These datasets were meticulously acquired in a simulated environment, necessitating the creation of several custom datasets. This research highlights the potential of deep ensemble classification in improving driver emotion recognition, thereby contributing to enhanced road safety.
{"title":"Enhancing Driver Emotion Recognition Through Deep Ensemble Classification","authors":"Faizan Zaman;Zhigang Xu;Adil Hussain;Anees Ullah;Khalid Zaman","doi":"10.26599/JICV.2025.9210055","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210055","url":null,"abstract":"This research addresses the challenging task of classifying drivers' emotions to increase their awareness of their driving behaviors. It recognizes the common issue of driver emotions, which often leads to the neglect of poor driving practices. By automatically detecting and identifying these behaviors, drivers can proactively obtain valuable insights to reduce potential accidents. This study proposes a comprehensive facial recognition model for drivers that uses a unified architecture comprising a convolutional neural network (CNN), a recurrent neural network (RNN), and a multilayer perceptron (MLP) classification model. Initially, a faster region-based convolutional neural network (R-CNN) was employed for accurate and efficient facial detection of drivers in live and recorded videos. Features are extracted from three CNN models and merged via advanced techniques to create an ensemble classification model. Moreover, the improved Faster R-CNN feature learning module is replaced with a new convolutional neural network module, VGG16, which maximizes the precision and effectiveness of facial detection in our system. Significant accuracy results of 89.2%, 97.20%, 99.01%, 93.65%, and 98.61% are shown in evaluations of our suggested facial detection and facial expression recognition (DFER) datasets, including the EMOTIC, CK+, FERPLUS, AffectNet, and custom datasets. These datasets were meticulously acquired in a simulated environment, necessitating the creation of several custom datasets. This research highlights the potential of deep ensemble classification in improving driver emotion recognition, thereby contributing to enhanced road safety.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210055-1-9210055-16"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646570","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the development of autonomous driving technology, traffic mixed with human-driven vehicles (HDVs) and autonomous vehicles (AVs) has dominated transportation systems for a long period of time. Drivers' car-following decision-making in mixed traffic needs to be considered for traffic simulation and management policy formulation. This study aims to explore the differences in drivers' decision-making mechanisms when following AVs and HDVs. Data from a questionnaire survey and a field test are collected and employed to establish a Bayesian network for car-following decision-making process analysis and inference. The influences of driving habits and recognition of AVs on car-following decisions and the correlations among the four decision variables are analyzed. The four decision variables consist of the vehicle gap and acceleration in both the acceleration and deceleration phases. The results show that there are direct correlations among the four internal decision variables. Among the external variables, overspeeding and honking have distinct impacts on decisions made while following an AV. Moreover, regardless of whether they are in an acceleration or deceleration phase, most drivers tend to make gentler decisions when following AVs than when following HDVs. On the basis of the results, we propose some strategies for the traffic management of mixed traffic that are beneficial to traffic efficiency: (1) Improving drivers' recognition of AVs; (2) embedding the external sensing devices of AVs internally to make them visually similar to HDVs; and (3) establishing dedicated lanes for AVs. The research results have important reference significance for simulating car-following behavior, designing traffic control facilities and formulating policies under mixed traffic scenarios.
{"title":"Decision-Making of Drivers Following Autonomous Vehicles: Developing a Bayesian Network on the Basis of Field Tests and Questionnaire Data","authors":"Fang Zong;Huan Wu;Meng Zeng;Won Kim;Qiaowen Bai;Yafeng Gong;Ruifeng Duan;Ying Guo","doi":"10.26599/JICV.2025.9210057","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210057","url":null,"abstract":"With the development of autonomous driving technology, traffic mixed with human-driven vehicles (HDVs) and autonomous vehicles (AVs) has dominated transportation systems for a long period of time. Drivers' car-following decision-making in mixed traffic needs to be considered for traffic simulation and management policy formulation. This study aims to explore the differences in drivers' decision-making mechanisms when following AVs and HDVs. Data from a questionnaire survey and a field test are collected and employed to establish a Bayesian network for car-following decision-making process analysis and inference. The influences of driving habits and recognition of AVs on car-following decisions and the correlations among the four decision variables are analyzed. The four decision variables consist of the vehicle gap and acceleration in both the acceleration and deceleration phases. The results show that there are direct correlations among the four internal decision variables. Among the external variables, overspeeding and honking have distinct impacts on decisions made while following an AV. Moreover, regardless of whether they are in an acceleration or deceleration phase, most drivers tend to make gentler decisions when following AVs than when following HDVs. On the basis of the results, we propose some strategies for the traffic management of mixed traffic that are beneficial to traffic efficiency: (1) Improving drivers' recognition of AVs; (2) embedding the external sensing devices of AVs internally to make them visually similar to HDVs; and (3) establishing dedicated lanes for AVs. The research results have important reference significance for simulating car-following behavior, designing traffic control facilities and formulating policies under mixed traffic scenarios.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210057-1-9210057-18"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646574","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01DOI: 10.26599/JICV.2025.9210063
Linsong Xue;Qi Luo;Kai Zhang
Localization along fixed routes is the fundamental function of transportation applications, including patrol vehicles, shuttles, buses, and even passenger vehicles. To achieve accurate and reliable localization, we propose a tightly coupled A Priori Map Simultaneous Localization and Mapping (APM-SLAM) system. APM-SLAM provides a comprehensive and heterogeneous framework, encompassing both mapping and localization processes. The mapping stage leverages Global Navigation Satellite System (GNSS)-aided Structure from Motion (SfM) to establish reliable a priori maps with coarse-and fine-level components. The localization process integrates coarse-to-fine matching with Maximum A Posteriori (MAP) Probability estimation to refine pose accuracy. By incorporating deep learning-based features and point descriptors, our system maintains robustness even in scenarios with significant visual variation. Unlike traditional map-based approaches, APM-SLAM models the a priori map's point structures as probabilistic distributions and incorporates them into the optimization process. Extensive experiments on public datasets demonstrate the superiority of our method in both mapping precision and localization accuracy, achieving decimeter-level translation precision. Ablation studies further validate the effectiveness of each component within our system. This work contributes to establishing maps and utilizing a priori information for localization simultaneously.
{"title":"APM-SLAM: Visual Localization for Fixed Routes with Tightly Coupled a Priori Map","authors":"Linsong Xue;Qi Luo;Kai Zhang","doi":"10.26599/JICV.2025.9210063","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210063","url":null,"abstract":"Localization along fixed routes is the fundamental function of transportation applications, including patrol vehicles, shuttles, buses, and even passenger vehicles. To achieve accurate and reliable localization, we propose a tightly coupled A Priori Map Simultaneous Localization and Mapping (APM-SLAM) system. APM-SLAM provides a comprehensive and heterogeneous framework, encompassing both mapping and localization processes. The mapping stage leverages Global Navigation Satellite System (GNSS)-aided Structure from Motion (SfM) to establish reliable a priori maps with coarse-and fine-level components. The localization process integrates coarse-to-fine matching with Maximum A Posteriori (MAP) Probability estimation to refine pose accuracy. By incorporating deep learning-based features and point descriptors, our system maintains robustness even in scenarios with significant visual variation. Unlike traditional map-based approaches, APM-SLAM models the a priori map's point structures as probabilistic distributions and incorporates them into the optimization process. Extensive experiments on public datasets demonstrate the superiority of our method in both mapping precision and localization accuracy, achieving decimeter-level translation precision. Ablation studies further validate the effectiveness of each component within our system. This work contributes to establishing maps and utilizing a priori information for localization simultaneously.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210063-1-9210063-15"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01DOI: 10.26599/JICV.2025.9210059
Zhiqiang Hu;Mingxing Xu;Qixiu Cheng
Autonomous driving technology has made significant advancements in recent years. The evolution of autonomous driving systems from traditional modular designs to end-to-end learning paradigms has led to comprehensive improvements in driving capabilities. In modular designs, driving tasks are segmented into independent modules, such as perception, decision-making, planning, and control. This modular structure offers high explainability and safety in simple scenarios but is hindered by limited generalizability in complex traffic environments, and the sequential connection of multiple modules often leads to error accumulation. In contrast, end-to-end methods process perception data directly to produce control outputs, thereby mitigating information loss and sequential error accumulation, ultimately improving scene generalization in diverse environments. However, this approach is limited by strong data dependency, low interpretability, and inadequate handling of long-tail scenarios (Zhao et al., 2024).
近年来，自动驾驶技术取得了重大进展。自动驾驶系统从传统的模块化设计向端到端学习范式的演变，导致了驾驶能力的全面提高。在模块化设计中，驾驶任务被分割成独立的模块，如感知、决策、规划和控制。这种模块化结构在简单场景下具有较高的可解释性和安全性，但在复杂交通环境下泛化能力有限，且多个模块的顺序连接往往导致错误积累。相比之下，端到端方法直接处理感知数据以产生控制输出，从而减少信息丢失和顺序误差积累，最终提高不同环境下的场景泛化。然而，这种方法受到数据依赖性强、可解释性低以及对长尾场景处理不足的限制（Zhao et al., 2024）。
{"title":"Multimodal Large-Language Model Empowering Next-Generation Autonomous Driving Systems","authors":"Zhiqiang Hu;Mingxing Xu;Qixiu Cheng","doi":"10.26599/JICV.2025.9210059","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210059","url":null,"abstract":"Autonomous driving technology has made significant advancements in recent years. The evolution of autonomous driving systems from traditional modular designs to end-to-end learning paradigms has led to comprehensive improvements in driving capabilities. In modular designs, driving tasks are segmented into independent modules, such as perception, decision-making, planning, and control. This modular structure offers high explainability and safety in simple scenarios but is hindered by limited generalizability in complex traffic environments, and the sequential connection of multiple modules often leads to error accumulation. In contrast, end-to-end methods process perception data directly to produce control outputs, thereby mitigating information loss and sequential error accumulation, ultimately improving scene generalization in diverse environments. However, this approach is limited by strong data dependency, low interpretability, and inadequate handling of long-tail scenarios (Zhao et al., 2024).","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210059-1-9210059-3"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11083713","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646573","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01DOI: 10.26599/JICV.2025.9210060
Bo Zhang;Heye Huang;Chunyang Liu;Yaqin Zhang;Zhenhua Xu
End-to-end autonomous driving, with its holistic optimization capabilities, has gained increasing traction in academia and industry. Vectorized representations, which preserve instance-level topological information while reducing computational overhead, have emerged as promising paradigms. However, existing vectorized query-based frameworks often overlook the inherent spatial correlations among intra-instance points, resulting in geometrically inconsistent outputs (e.g., fragmented HD map elements or oscillatory trajectories). To address these limitations, we propose intra-instance vectorized driving transformer (InVDriver), a novel vectorized query-based system that systematically models intra-instance spatial dependencies through masked self-attention layers, thereby enhancing planning accuracy and trajectory smoothness. Across all core modules, i.e., perception, prediction, and planning, InVDriver incorporates masked self-attention mechanisms that restrict attention to intra-instance point interactions, enabling coordinated refinement of structural elements while suppressing irrelevant inter-instance noise. The experimental results on the nuScenes benchmark demonstrate that InVDriver achieves state-of-the-art performance, surpassing prior methods in both accuracy and safety, while maintaining high computational efficiency.
{"title":"InVDriver: Intra-Instance Aware Vectorized Query-Based Autonomous Driving Transformer","authors":"Bo Zhang;Heye Huang;Chunyang Liu;Yaqin Zhang;Zhenhua Xu","doi":"10.26599/JICV.2025.9210060","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210060","url":null,"abstract":"End-to-end autonomous driving, with its holistic optimization capabilities, has gained increasing traction in academia and industry. Vectorized representations, which preserve instance-level topological information while reducing computational overhead, have emerged as promising paradigms. However, existing vectorized query-based frameworks often overlook the inherent spatial correlations among intra-instance points, resulting in geometrically inconsistent outputs (e.g., fragmented HD map elements or oscillatory trajectories). To address these limitations, we propose intra-instance vectorized driving transformer (InVDriver), a novel vectorized query-based system that systematically models intra-instance spatial dependencies through masked self-attention layers, thereby enhancing planning accuracy and trajectory smoothness. Across all core modules, i.e., perception, prediction, and planning, InVDriver incorporates masked self-attention mechanisms that restrict attention to intra-instance point interactions, enabling coordinated refinement of structural elements while suppressing irrelevant inter-instance noise. The experimental results on the nuScenes benchmark demonstrate that InVDriver achieves state-of-the-art performance, surpassing prior methods in both accuracy and safety, while maintaining high computational efficiency.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210060-1-9210060-8"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646572","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01DOI: 10.26599/JICV.2025.9210056
Jiajie Zhang;Bao-Lin Ye;Xin Wang;Lingxi Li;Bo Song
To improve the driving efficiency of unmanned vehicles in a complex urban traffic flow environment and the safety and passenger comfort of vehicles when changing lanes, we propose a hierarchical reinforcement learning (HRL)-based vehicle trajectory planning and tracking method. First, we present a hierarchical control framework for vehicle trajectory tracking that is based on deep reinforcement learning (DRL) and model predictive control (MPC). We design an upper-level decision model based on the trust region policy optimization algorithm integrated with long short-term memory to obtain more accurate strategies. Second, to improve stability and passenger comfort, we constructed a lower controller that combines the Bezier curve fitting method and an MPC controller. Finally, the proposed method was simulated via the car learning to act (CARLA) simulator, which is based on an unreal engine. Random urban traffic-flow test scenarios were used to simulate a real urban road-traffic environment. The simulation results illustrate that the proposed method can complete the vehicle trajectory planning and tracking task well. Compared with the existing RL methods, our proposed method has the lowest collision rate of 1.5% and achieves an average speed improvement of 7.04%. Moreover, our proposed method has better comfort performance and lower fuel consumption during the driving process.
{"title":"A Trajectory Planning and Tracking Method Based on Deep Hierarchical Reinforcement Learning","authors":"Jiajie Zhang;Bao-Lin Ye;Xin Wang;Lingxi Li;Bo Song","doi":"10.26599/JICV.2025.9210056","DOIUrl":"https://doi.org/10.26599/JICV.2025.9210056","url":null,"abstract":"To improve the driving efficiency of unmanned vehicles in a complex urban traffic flow environment and the safety and passenger comfort of vehicles when changing lanes, we propose a hierarchical reinforcement learning (HRL)-based vehicle trajectory planning and tracking method. First, we present a hierarchical control framework for vehicle trajectory tracking that is based on deep reinforcement learning (DRL) and model predictive control (MPC). We design an upper-level decision model based on the trust region policy optimization algorithm integrated with long short-term memory to obtain more accurate strategies. Second, to improve stability and passenger comfort, we constructed a lower controller that combines the Bezier curve fitting method and an MPC controller. Finally, the proposed method was simulated via the car learning to act (CARLA) simulator, which is based on an unreal engine. Random urban traffic-flow test scenarios were used to simulate a real urban road-traffic environment. The simulation results illustrate that the proposed method can complete the vehicle trajectory planning and tracking task well. Compared with the existing RL methods, our proposed method has the lowest collision rate of 1.5% and achieves an average speed improvement of 7.04%. Moreover, our proposed method has better comfort performance and lower fuel consumption during the driving process.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 2","pages":"9210056-1-9210056-9"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-01DOI: 10.26599/JICV.2024.9210053
Donglei Rong;Yuefeng Wu;Wenjun Du;Chengcheng Yang;Sheng Jin;Min Xu;Fujian Wang
To improve the adaptability of Connected and Automated Vehicles (CAVs) in mixed traffic, this study proposes a prediction model training indicator that comprehensively considers drivers' Social Value Orientation (SVO) and planning goals. Active Influence Factor (AIF) is used as the goal to predict the future safety loss and consistency loss of CAVs. Second, an objective function based on SVO is constructed to understand the driver's characteristics to evaluate the safety, comfort, efficiency, and consistency of candidate trajectories. The results showed that integrating SVO and consistency functions can help ensure that CAVs drive under a more stable risk potential energy field. The prediction planning model that considers SVO can improve the reliability of the CAV output trajectory to a certain extent. The prediction planning under the AIF has better accuracy and stability of the output trajectory; however, it still has strong adaptability and superiority under different sensitivity parameters. The minimum and maximum standard deviations of our model are 0.78 and 0.78 m, respectively, whereas the minimum and maximum standard deviations of the comparative model reach 2.07 and 4.56 m, respectively. The minimum standard deviation of the other comparative model reaches 1.35 m, and the maximum standard deviation reaches 4.45 m.
{"title":"Smart Prediction-Planning Algorithm for Connected and Autonomous Vehicle Based on Social Value Orientation","authors":"Donglei Rong;Yuefeng Wu;Wenjun Du;Chengcheng Yang;Sheng Jin;Min Xu;Fujian Wang","doi":"10.26599/JICV.2024.9210053","DOIUrl":"https://doi.org/10.26599/JICV.2024.9210053","url":null,"abstract":"To improve the adaptability of Connected and Automated Vehicles (CAVs) in mixed traffic, this study proposes a prediction model training indicator that comprehensively considers drivers' Social Value Orientation (SVO) and planning goals. Active Influence Factor (AIF) is used as the goal to predict the future safety loss and consistency loss of CAVs. Second, an objective function based on SVO is constructed to understand the driver's characteristics to evaluate the safety, comfort, efficiency, and consistency of candidate trajectories. The results showed that integrating SVO and consistency functions can help ensure that CAVs drive under a more stable risk potential energy field. The prediction planning model that considers SVO can improve the reliability of the CAV output trajectory to a certain extent. The prediction planning under the AIF has better accuracy and stability of the output trajectory; however, it still has strong adaptability and superiority under different sensitivity parameters. The minimum and maximum standard deviations of our model are 0.78 and 0.78 m, respectively, whereas the minimum and maximum standard deviations of the comparative model reach 2.07 and 4.56 m, respectively. The minimum standard deviation of the other comparative model reaches 1.35 m, and the maximum standard deviation reaches 4.45 m.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 1","pages":"1-17"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10960599","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143808988","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-01DOI: 10.26599/JICV.2024.9210054
Ling Han;Xiangyu Ma;Yiren Wang;Lei He;Yipeng Li;Lele Zhang;Qiang Yi
Lane changing is common in driving. Thus, the possibility of traffic accidents occurring during lane changes is high given the complexity of this process. One of the primary objectives of intelligent driving is to increase a vehicle's behavior, making it more similar to that of a real driver. This study proposes a decision-making framework based on deep reinforcement learning (DRL) in a lane-changing scenario, which seeks to find a driving strategy that simultaneously considers the expected lane-changing risks and gains. First, a fuzzy logic lane-changing controller is designed. It outputs the corresponding safety and lane-change gain weights by inputting relevant driving parameters. Second, the obtained weights are brought into the constructed reward function of DRL. The model parameters are designed and trained on the basis of lane-changing behavior. Finally, we conducted experiments in a simulator to evaluate the performance of our developed algorithm in urban scenarios. To visualize and validate the estimated driving intentions, lane-changing strategies were tested under four scenarios. The results show that the average improvement in travel efficiency in the four scenarios is 19%. In addition, the average accident rate in the four scenarios increased by only 4%. We combine fuzzy logic and DRL reward functions to personify the lane-changing behavior of intelligent driving. Compared with conservative strategies that prioritize only safety, this method can considerably improve the number of lane changes and travel efficiency for autonomous vehicles (AVs) on the premise of ensuring safety. The approach provides an effective and explainable method designed for facilitating intelligent driving lane-changing behavior.
{"title":"Safe and Efficient DRL Driving Policies Using Fuzzy Logic for Urban Lane Changing Scenarios","authors":"Ling Han;Xiangyu Ma;Yiren Wang;Lei He;Yipeng Li;Lele Zhang;Qiang Yi","doi":"10.26599/JICV.2024.9210054","DOIUrl":"https://doi.org/10.26599/JICV.2024.9210054","url":null,"abstract":"Lane changing is common in driving. Thus, the possibility of traffic accidents occurring during lane changes is high given the complexity of this process. One of the primary objectives of intelligent driving is to increase a vehicle's behavior, making it more similar to that of a real driver. This study proposes a decision-making framework based on deep reinforcement learning (DRL) in a lane-changing scenario, which seeks to find a driving strategy that simultaneously considers the expected lane-changing risks and gains. First, a fuzzy logic lane-changing controller is designed. It outputs the corresponding safety and lane-change gain weights by inputting relevant driving parameters. Second, the obtained weights are brought into the constructed reward function of DRL. The model parameters are designed and trained on the basis of lane-changing behavior. Finally, we conducted experiments in a simulator to evaluate the performance of our developed algorithm in urban scenarios. To visualize and validate the estimated driving intentions, lane-changing strategies were tested under four scenarios. The results show that the average improvement in travel efficiency in the four scenarios is 19%. In addition, the average accident rate in the four scenarios increased by only 4%. We combine fuzzy logic and DRL reward functions to personify the lane-changing behavior of intelligent driving. Compared with conservative strategies that prioritize only safety, this method can considerably improve the number of lane changes and travel efficiency for autonomous vehicles (AVs) on the premise of ensuring safety. The approach provides an effective and explainable method designed for facilitating intelligent driving lane-changing behavior.","PeriodicalId":100793,"journal":{"name":"Journal of Intelligent and Connected Vehicles","volume":"8 1","pages":"1-13"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10960600","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143808987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：