Showing papers in "IEEE transactions on intelligent vehicles in 2023"
TL;DR: In this article , the authors present a survey of surveys for autonomous driving and intelligent vehicles (IVs) that reviews the history, summarizes the milestones, and provides the perspectives, ethics, and future research directions.
Abstract: Interest in autonomous driving (AD) and intelligent vehicles (IVs) is growing at a rapid pace due to the convenience, safety, and economic benefits. Although a number of surveys have reviewed research achievements in this field, they are still limited in specific tasks, lack of systematic summary and research directions in the future. Here we propose a Survey of Surveys (SoS) for total technologies of AD and IVs that reviews the history, summarizes the milestones, and provides the perspectives, ethics, and future research directions. To our knowledge, this article is the first SoS with milestones in AD and IVs, which constitutes our complete research work together with two other technical surveys. We anticipate that this article will bring novel and diverse insights to researchers and abecedarians, and serve as a bridge between past and future.
28 citations
TL;DR: Zhang et al. as mentioned in this paper proposed a contrastive learning approach with the aim of building a model that can quantify driver anomalies with a continuous variable and proposed a novel clustering supervised contrastive loss to optimize the distribution of the extracted representation vectors to improve the model performance.
Abstract: Driver anomaly quantification is a fundamental capability to support human-centric driving systems of intelligent vehicles. Existing studies usually treat it as a classification task and obtain discrete levels for abnormalities. Meanwhile, the existing data-driven approaches depend on the quality of dataset and provide limited recognition capability for unknown activities. To overcome these challenges, this paper proposes a contrastive learning approach with the aim of building a model that can quantify driver anomalies with a continuous variable. In addition, a novel clustering supervised contrastive loss is proposed to optimize the distribution of the extracted representation vectors to improve the model performance. Compared with the typical contrastive loss, the proposed loss can better cluster normal representations while separating abnormal ones. The abnormality of driver activity can be quantified by calculating the distance to a set of representations of normal activities rather than being produced as the direct output of the model. The experiment results with datasets under different modes demonstrate that the proposed approach is more accurate and robust than existing ones in terms of recognition and quantification of unknown abnormal activities.
27 citations
TL;DR: In this article , a constrained observation-robust Markov decision process is presented to model lane change decision making behaviors of autonomous vehicles under policy constraints and observation uncertainties, and a black-box attack technique based on Bayesian optimization is implemented to approximate the optimal adversarial observation perturbations efficiently.
Abstract: Reinforcementlearning holds the promise of allowing autonomous vehicles to learn complex decision making behaviors through interacting with other traffic participants. However, many real-world driving tasks involve unpredictable perception errors or measurement noises which may mislead an autonomous vehicle into making unsafe decisions, even cause catastrophic failures. In light of these risks, to ensure safety under perception uncertainty, autonomous vehicles are required to be able to cope with the worst case observation perturbations. Therefore, this paper proposes a novel observation adversarial reinforcement learning approach for robust lane change decision making of autonomous vehicles. A constrained observation-robust Markov decision process is presented to model lane change decision making behaviors of autonomous vehicles under policy constraints and observation uncertainties. Meanwhile, a black-box attack technique based on Bayesian optimization is implemented to approximate the optimal adversarial observation perturbations efficiently. Furthermore, a constrained observation-robust actor-critic algorithm is advanced to optimize autonomous driving lane change policies while keeping the variations of the policies attacked by the optimal adversarial observation perturbations within bounds. Finally, the robust lane change decision making approach is evaluated in three stochastic mixed traffic flows based on different densities. The results demonstrate that the proposed method can not only enhance the performance of an autonomous vehicle but also improve the robustness of lane change policies against adversarial observation perturbations.
27 citations
TL;DR: In this article , a multitask-aware network (MTANet) with hierarchical multimodal fusion (multiscale fusion strategy) was proposed for RGB-T urban scene understanding.
Abstract: Understanding urban scenes is a fundamental ability requirement for assisted driving and autonomous vehicles. Most of the available urban scene understanding methods use red-green-blue (RGB) images; however, their segmentation performances are prone to degradation under adverse lighting conditions. Recently, many effective artificial neural networks have been presented for urban scene understanding and have shown that incorporating RGB and thermal (RGB-T) images can improve segmentation accuracy even under unsatisfactory lighting conditions. However, the potential of multimodal feature fusion has not been fully exploited because operations such as simply concatenating the RGB and thermal features or averaging their maps have been adopted. To improve the fusion of multimodal features and the segmentation accuracy, we propose a multitask-aware network (MTANet) with hierarchical multimodal fusion (multiscale fusion strategy) for RGB-T urban scene understanding. We developed a hierarchical multimodal fusion module to enhance feature fusion and built a high-level semantic module to extract semantic information for merging with coarse features at various abstraction levels. Using the multilevel fusion module, we exploited low-, mid-, and high-level fusion to improve segmentation accuracy. The multitask module uses boundary, binary, and semantic supervision to optimize the MTANet parameters. Extensive experiments were performed on two benchmark RGB-T datasets to verify the improved performance of the proposed MTANet compared with state-of-the-art methods.1
19 citations
TL;DR: In this paper , a path planning method based on double deep Q Network (DDQN) was proposed to improve the AUV's path planning capability in the unknown environments, which is created from an improved convolutional neural network, which has two input layers to adapt to the processing of high-dimensional environments.
Abstract: The path planning issue of the underactuated autonomous underwater vehicle (AUV) under ocean current disturbance is studied in this paper. In order to improve the AUV’s path planning capability in the unknown environments, a deep reinforcement learning (DRL) path planning method based on double deep Q Network (DDQN) is proposed. It is created from an improved convolutional neural network, which has two input layers to adapt to the processing of high-dimensional environments. Considering the maneuverability of underactuated AUV under current disturbance, especially, the issue of ocean current disturbance under unknown environments, a dynamic and composite reward function is developed to enable the AUV to reach the destination with obstacle avoidance. Finally, the path planning ability of the proposed method in the unknown environments is validated by simulation analysis and comparison studies.
18 citations
TL;DR: The OpenCDA research ecosystem as discussed by the authors is an open-source ecosystem for cooperative driving automata (CDA), which includes a model zoo, a suite of driving simulators at various resolutions, complete development toolkits for benchmark training/testing, and a scenario database/generator.
Abstract: Advances in Single-vehicle intelligence of automated driving has encountered great challenges because of limited capabilities in perception and interaction with complex traffic environments. Cooperative Driving Automation (CDA) has been considered a pivotal solution to next-generation automated driving and smart transportation. Though CDA has attracted much attention from both academia and industry, exploration of its potential is still in its infancy. In industry, companies tend to build their in-house data collection pipeline and research tools to tailor their needs and protect intellectual properties. Reinventing the wheels, however, wastes resources and limits the generalizability of the developed approaches since no standardized benchmarks exist. On the other hand, in academia, due to the absence of real-world traffic data and computation resources, researchers often investigate CDA topics in simplified and mostly simulated environments, restricting the possibility of scaling the research outputs to real-world scenarios. Therefore, there is an urgent need to establish an open-source ecosystem (OSE) to address the demands of different communities for CDA research, particularly in the early exploratory research stages, and provide the bridge to ensure an integrated development and testing pipeline that diverse communities can share. In this paper, we introduce the OpenCDA research ecosystem, a unified OSE integrated with a model zoo, a suite of driving simulators at various resolutions, large-scale real-world and simulated datasets, complete development toolkits for benchmark training/testing, and a scenario database/generator. We also demonstrate the effectiveness of OpenCDA OSE through example use cases, including cooperative 3D LiDAR detection, cooperative merge, cooperative camera-based map prediction, and adversarial scenario generation.
16 citations
TL;DR: In this article , the authors present a survey of surveys for autonomous driving and intelligent vehicles (IVs) that reviews the history, summarizes the milestones, and provides the perspectives, ethics, and future research directions.
Abstract: Interest in autonomous driving (AD) and intelligent vehicles (IVs) is growing at a rapid pace due to the convenience, safety, and economic benefits. Although a number of surveys have reviewed research achievements in this field, they are still limited in specific tasks, lack of systematic summary and research directions in the future. Here we propose a Survey of Surveys (SoS) for total technologies of AD and IVs that reviews the history, summarizes the milestones, and provides the perspectives, ethics, and future research directions. To our knowledge, this article is the first SoS with milestones in AD and IVs, which constitutes our complete research work together with two other technical surveys. We anticipate that this article will bring novel and diverse insights to researchers and abecedarians, and serve as a bridge between past and future.
16 citations
TL;DR: In this paper , an uncertainty-aware model-based reinforcement learning (RL) method is proposed and validated in autonomous driving scenarios, where an action-conditioned ensemble model with the capability of uncertainty assessment is established as the environment model.
Abstract: To further improve learning efficiency and performance of reinforcement learning (RL), a novel uncertainty-aware model-based RL method is proposed and validated in autonomous driving scenarios in this paper. First, an action-conditioned ensemble model with the capability of uncertainty assessment is established as the environment model. Then, a novel uncertainty-aware model-based RL method is developed based on the adaptive truncation approach, providing virtual interactions between the agent and environment model, and improving RL’s learning efficiency and performance. The proposed method is then implemented in end-to-end autonomous vehicle control tasks, validated and compared with state-of-the-art methods under various driving scenarios. Validation results suggest that the proposed method outperforms the model-free RL approach with respect to learning efficiency, and model-based approach with respect to both efficiency and performance, demonstrating its feasibility and effectiveness.
16 citations
TL;DR: Zhang et al. as discussed by the authors proposed an Attention-based Interaction-aware Trajectory Prediction (AI-TP) model using Graph Attention Networks (GAT) to describe the interactions of traffic agents and Convolutional Gated Recurrent Units (ConvGRU) to carry out predictions.
Abstract: Despite the advancements in the technologies of autonomous driving, it is still challenging to study the safety of a self-driving vehicle. Trajectory prediction is one core function of an autonomous vehicle. This study proposes an Attention-based Interaction-aware Trajectory Prediction (AI-TP) for traffic agents around the autonomous vehicle. With an encoder-decoder architecture, the AI-TP model uses Graph Attention Networks (GAT) to describe the interactions of traffic agents and Convolutional Gated Recurrent Units (ConvGRU) to carry out predictions. Based on the attention mechanism, the AI-TP model constructs graphs from various traffic scenes to predict trajectories of different types of traffic agents. Traffic data from both the high-way (i.e., NGSIM) and urban road areas (i.e., ApolloScape and Argoverse) are used to evaluate the performance of the AI-TP model. Numerical results demonstrate that the AI-TP model requires less inference time and achieves better prediction accuracy than state-of-the-art methods. Specifically, the AI-TP model improves the performance with much less inference time on the NGSIM dataset, which shows the promise of predicting trajectories under various scenarios. The code of the AI-TP model will be available at https://github.com/KP-Zhang/AI-TP.
16 citations
TL;DR: Zhang et al. as discussed by the authors proposed Hierarchical Interpretable Imitation Learning (HIIL) model, which integrates interpretable BEV mask and steering angle to solve the problems of stability and interpretability in complex urban scenarios.
Abstract: End-to-end autonomous driving provides a simple and efficient framework for autonomous driving systems, which can directly obtain control commands from raw perception data. However, it fails to address stability and interpretability problems in complex urban scenarios. In this paper, we construct a two-stage end-to-end autonomous driving model for complex urban scenarios, named HIIL (Hierarchical Interpretable Imitation Learning), which integrates interpretable BEV mask and steering angle to solve the problems shown above. In Stage One, we propose a pretrained Bird's Eye View (BEV) model which leverages a BEV mask to present an interpretation of the surrounding environment. In Stage Two, we construct an Interpretable Imitation Learning (IIL) model that fuses BEV latent feature from Stage One with an additional steering angle from Pure-Pursuit (PP) algorithm. In the HIIL model, visual information is converted to semantic images by the semantic segmentation network, and the semantic images are encoded to extract the BEV latent feature, which are decoded to predict BEV masks and fed to the IIL as perception data. In this way, the BEV latent feature bridges the BEV and IIL models. Visual information can be supplemented by the calculated steering angle for PP algorithm, speed vector, and location information, thus it could have better performance in complex and terrible scenarios. Our HIIL model meets an urgent requirement for interpretability and robustness of autonomous driving. We validate the proposed model in the CARLA simulator with extensive experiments which show remarkable interpretability, generalization, and robustness capability in unknown scenarios for navigation tasks.
14 citations
TL;DR: In this paper , a constrained observation-robust Markov decision process is presented to model lane change decision making behaviors of autonomous vehicles under policy constraints and observation uncertainties, and a black-box attack technique based on Bayesian optimization is implemented to approximate the optimal adversarial observation perturbations efficiently.
Abstract: Reinforcementlearning holds the promise of allowing autonomous vehicles to learn complex decision making behaviors through interacting with other traffic participants. However, many real-world driving tasks involve unpredictable perception errors or measurement noises which may mislead an autonomous vehicle into making unsafe decisions, even cause catastrophic failures. In light of these risks, to ensure safety under perception uncertainty, autonomous vehicles are required to be able to cope with the worst case observation perturbations. Therefore, this paper proposes a novel observation adversarial reinforcement learning approach for robust lane change decision making of autonomous vehicles. A constrained observation-robust Markov decision process is presented to model lane change decision making behaviors of autonomous vehicles under policy constraints and observation uncertainties. Meanwhile, a black-box attack technique based on Bayesian optimization is implemented to approximate the optimal adversarial observation perturbations efficiently. Furthermore, a constrained observation-robust actor-critic algorithm is advanced to optimize autonomous driving lane change policies while keeping the variations of the policies attacked by the optimal adversarial observation perturbations within bounds. Finally, the robust lane change decision making approach is evaluated in three stochastic mixed traffic flows based on different densities. The results demonstrate that the proposed method can not only enhance the performance of an autonomous vehicle but also improve the robustness of lane change policies against adversarial observation perturbations.
TL;DR: In this paper , a path planning method based on double deep Q Network (DDQN) was proposed to improve the AUV's path planning capability in the unknown environments, which is created from an improved convolutional neural network, which has two input layers to adapt to the processing of high-dimensional environments.
Abstract: The path planning issue of the underactuated autonomous underwater vehicle (AUV) under ocean current disturbance is studied in this paper. In order to improve the AUV’s path planning capability in the unknown environments, a deep reinforcement learning (DRL) path planning method based on double deep Q Network (DDQN) is proposed. It is created from an improved convolutional neural network, which has two input layers to adapt to the processing of high-dimensional environments. Considering the maneuverability of underactuated AUV under current disturbance, especially, the issue of ocean current disturbance under unknown environments, a dynamic and composite reward function is developed to enable the AUV to reach the destination with obstacle avoidance. Finally, the path planning ability of the proposed method in the unknown environments is validated by simulation analysis and comparison studies.
TL;DR: In this paper , the longitudinal control of vehicle platoons with a focus on external disturbances, parameter uncertainties, and communication failures is investigated, where a generalized extended state observer is designed to estimate the lumped disturbance and the preceding vehicle's acceleration separately.
Abstract: This paper investigates the longitudinal control of vehicle platoons with a focus on external disturbances, parameter uncertainties, and communication failures. When vehicle platoons encounter vehicle to vehicle wireless communication failures, the preceding vehicle’s acceleration is unavailable, which degrades the performance of the system. However, the existing achievements can not be directly used to address the aforementioned three issues. To this end, a generalized extended state observer-based control (GESOBC) law is contrived. First, the parameter uncertainties and the external disturbances are together regarded as a lumped disturbance. Then, a generalized extended state observer is designed to estimate the lumped disturbance and the preceding vehicle’s acceleration, separately. Based on the estimation, a composite controller consisting of a state feedback control part and an estimation-based feedforward control part is developed. Furthermore, it is proved that the proposed GESOBC method can guarantee the exponentially bounded stability of the individual vehicle and the input to state string stability of the whole vehicle platoon. Finally, numerical simulations are conducted to demonstrate the effectiveness and feasibility of the proposed method.
TL;DR: In this article , an event-triggered estimation framework by fusing an event triggered mechanism with an embedded cubature Kalman filter based on a coupled vehicle model is proposed for PVs state estimation.
Abstract: Accurate knowledge about the motion state of preceding vehicles (PVs) contributes to the optimization of planning and decision making of autonomous vehicles, which in turn further enhances road safety. Existing studies generally rely on information from special sensors mounted on the ego vehicle to estimate PVs state. With the evolution of information technology, the use of vehicle-to-vehicle (V2V) communication to estimate the PVs state has attracted more and more attention. However, the problem of how to reduce the communication rate while ensuring the estimation accuracy of PVs state with the limited communication bandwidth has not been addressed yet. In addition, traditional studies on lateral state estimation of PVs assume that the longitudinal velocity of the PV is known or design an additional estimator to predict the longitudinal velocity. This results in the dynamical coupling characteristics of the vehicle not being fully considered. To address these problems, an event-triggered estimation framework by fusing an event-triggered mechanism with an embedded cubature Kalman filter based on a coupled vehicle model is proposed for PVs state estimation. Simulation and real vehicle test results demonstrate that the proposed prediction approach can strike an effective balance between the communication rate and the estimation performance. The estimation accuracy of the proposed method is still superior to that of the cubature Kalman filter, even if the communication rate drops to 37.55%.
TL;DR: In this article , the authors proposed a data-driven method to model mixed vehicle platoons based on Koopman operator theory, which gives a way to represent the mixed vehicle platoon by a linear model in a high-dimensional space, the approximation of which is obtained by a neural network framework.
Abstract: With the development of automatic driving technology and the internet of vehicles, platooning based on control of connected autonomous vehicles has become one of the most promising approaches to improve traffic efficiency. This paper studies the control problem of mixed vehicle platoons consisting of human-driven vehicles and connected autonomous vehicles. Firstly, we propose a data-driven method to model mixed vehicle platoons based on Koopman operator theory. This method gives a way to represent the mixed vehicle platoon by a linear model in a high-dimensional space, the approximation of which is obtained by a neural network framework. Secondly, we employ model predictive control (MPC) to address the platoon control problem of mixed vehicle platoons, where both centralized MPC and distributed MPC algorithms are designed. Finally, the effectiveness of the data-driven modeling method and the centralized/distributed MPC algorithms is verified by numerical simulations. It is revealed that the proposed data-driven DMPC algorithm exhibits comparable control performance with less computation cost compared with the centralized MPC algorithm, and it shows faster convergence speed than the nonlinear model based DMPC algorithm.
TL;DR: In this paper , state-of-the-art motion planning methods for intelligent vehicles, including pipeline planning and end-to-end planning methods, are reviewed to highlight their strengths and limitations.
Abstract: Intelligent vehicles (IVs) have gained worldwide attention due to their increased convenience, safety advantages, and potential commercial value. Despite predictions of commercial deployment by 2025, implementation remains limited to small-scale validation, with precise tracking controllers and motion planners being essential prerequisites for IVs. This article reviews state-of-the-art motion planning methods for IVs, including pipeline planning and end-to-end planning methods. The study examines the selection, expansion, and optimization operations in a pipeline method, while it investigates training approaches and validation scenarios for driving tasks in end-to-end methods. Experimental platforms are reviewed to assist readers in choosing suitable training and validation strategies. A side-by-side comparison of the methods is provided to highlight their strengths and limitations, aiding system-level design choices. Current challenges and future perspectives are also discussed in this survey.
TL;DR: In this article , a longitudinal and lateral controller is designed to enable a connected automated vehicle (CAV) to utilize V2X information from nearby connected human-driven vehicles (CHVs), while taking into account the time delays in the feedback loops.
Abstract: The concept of utilizing vehicle-to-everything (V2X) connectivity to improve the resilience of automated vehicles in an environment where optical sensors may not provide reliable data is investigated. Longitudinal and lateral controllers are designed to enable a connected automated vehicle (CAV) to utilize V2X information from nearby connected human-driven vehicles (CHVs). The linear stability of the controllers are investigated theoretically while taking into account the time delays in the feedback loops. Novel performance measures are introduced to quantify the plant and string stability properties of the longitudinal controller from experimental data. The stability of the lateral controller is also evaluated in lane-keeping experiments. The robustness of the designed controllers against latency is demonstrated and the performance of the overall systems is showcased experimentally using real vehicles on a test track.
TL;DR: In this paper , a novel framework called parallel vehicular crowd sensing (P-VCS) is proposed to balance the physical environment, cyber networks, and human and social factors.
Abstract: As an emerging paradigm for urban sensing, vehicular crowd sensing (VCS) has received increasing attention in recent years. Unlike traditional sensing paradigms, VCS leverages ubiquitous connected vehicles (CVs) and diverse onboard sensors to efficiently collect city-scale data. Despite the considerable benefits of CVs, the fast-changing traffic environment and attendant human and social factors bring significant complexity to the VCS system and make it a typical cyber-physical-social system (CPSS), followed by the challenge of robust and efficient modeling of VCS systems. To cope with the complexity of social dimensions and optimize the decision-making process in the physical VCS, this article introduces the artificial societies, computational experiments, and the parallel execution (ACP) approach to the VCS system and develops a novel framework called parallel VCS (P-VCS). Three key components empower P-VCS to balance the physical environment, cyber networks, and human and social factors, namely, an artificial system that is used to parametrically describe the physical VCS, two types of computational experiments that simulate the decision process and evaluate different strategies, and the parallel execution mechanism that is used to characterize the system operation. To demonstrate the feasibility of the framework, we take participant selection under traffic events as an application example. Experimental results illustrate that the P-VCS-based parallel learning strategy maintains competitive performance in all cases.
TL;DR: Zhang et al. as mentioned in this paper proposed Hierarchical Interpretable Imitation Learning (HIIL) model for complex urban scenarios, which integrates interpretable BEV mask and steering angle to solve the problems shown above.
Abstract: End-to-end autonomous driving provides a simple and efficient framework for autonomous driving systems, which can directly obtain control commands from raw perception data. However, it fails to address stability and interpretability problems in complex urban scenarios. In this paper, we construct a two-stage end-to-end autonomous driving model for complex urban scenarios, named HIIL (Hierarchical Interpretable Imitation Learning), which integrates interpretable BEV mask and steering angle to solve the problems shown above. In Stage One, we propose a pretrained Bird's Eye View (BEV) model which leverages a BEV mask to present an interpretation of the surrounding environment. In Stage Two, we construct an Interpretable Imitation Learning (IIL) model that fuses BEV latent feature from Stage One with an additional steering angle from Pure-Pursuit (PP) algorithm. In the HIIL model, visual information is converted to semantic images by the semantic segmentation network, and the semantic images are encoded to extract the BEV latent feature, which are decoded to predict BEV masks and fed to the IIL as perception data. In this way, the BEV latent feature bridges the BEV and IIL models. Visual information can be supplemented by the calculated steering angle for PP algorithm, speed vector, and location information, thus it could have better performance in complex and terrible scenarios. Our HIIL model meets an urgent requirement for interpretability and robustness of autonomous driving. We validate the proposed model in the CARLA simulator with extensive experiments which show remarkable interpretability, generalization, and robustness capability in unknown scenarios for navigation tasks.
TL;DR: In this paper , the authors explore the RSMA scheme for RIS-enabled UAV-based multi-user vehicular communication network in the presence of co-channel interference and derive the approximate cumulative distribution function (CDF) of common and private stream signal-to-interference-plus-noise ratios (SINRs) at the desired vehicle in the interference limited scenario, obtaining an expression for the average outage probability (AOP) over the complete flying duration of the UAV.
Abstract: Rate-splitting multiple access (RSMA) has emerged as a novel generalized multiple access technology which can control the interference in multi-user communication systems. Reconfigurable intelligent surface (RIS) has also been proven to improve the spectral efficiency in the next-generation (beyond 5 G (B5G) and 6 G) wireless communication networks. Thus, in this paper, we explore the RSMA scheme for RIS-enabled unmanned aerial vehicle (UAV) based multi-user vehicular communication network in the presence of co-channel interference. The communication between the UAV and the desired vehicle takes place through an RIS along with the direct link and is interfered by multiple interfering vehicles operating in the same spectrum. After deriving the approximate cumulative distribution function (CDF) of common and private stream signal-to-interference-plus-noise ratios (SINRs) at the desired vehicle in the interference limited scenario, we obtain an expression for the average outage probability (AOP) over the complete flying duration of the UAV. Furthermore, the optimum RSMA power allocation coefficients in each time slot of UAV flight duration are obtained by minimizing the Sum-AOP over all the desired vehicles. The performance of the proposed RSMA based RIS-aided vehicular communication network is compared with (a) RSMA based network without RIS and (b) non-orthogonal multiple access (NOMA) based RIS-aided counterpart. The numerical results reveal the superiority of RSMA over NOMA and the significance of deploying RIS in considered vehicular communication network. The analytical results are corroborated with Monte Carlo simulations.
TL;DR: In this article , the authors proposed a data-driven method to model mixed vehicle platoons based on Koopman operator theory, which gives a way to represent the mixed vehicle platoon by a linear model in a high-dimensional space, the approximation of which is obtained by a neural network framework.
Abstract: With the development of automatic driving technology and the internet of vehicles, platooning based on control of connected autonomous vehicles has become one of the most promising approaches to improve traffic efficiency. This paper studies the control problem of mixed vehicle platoons consisting of human-driven vehicles and connected autonomous vehicles. Firstly, we propose a data-driven method to model mixed vehicle platoons based on Koopman operator theory. This method gives a way to represent the mixed vehicle platoon by a linear model in a high-dimensional space, the approximation of which is obtained by a neural network framework. Secondly, we employ model predictive control (MPC) to address the platoon control problem of mixed vehicle platoons, where both centralized MPC and distributed MPC algorithms are designed. Finally, the effectiveness of the data-driven modeling method and the centralized/distributed MPC algorithms is verified by numerical simulations. It is revealed that the proposed data-driven DMPC algorithm exhibits comparable control performance with less computation cost compared with the centralized MPC algorithm, and it shows faster convergence speed than the nonlinear model based DMPC algorithm.
TL;DR: In this paper , a novel framework called parallel vehicular crowd sensing (P-VCS) is proposed to balance the physical environment, cyber networks, and human and social factors.
Abstract: As an emerging paradigm for urban sensing, vehicular crowd sensing (VCS) has received increasing attention in recent years. Unlike traditional sensing paradigms, VCS leverages ubiquitous connected vehicles (CVs) and diverse onboard sensors to efficiently collect city-scale data. Despite the considerable benefits of CVs, the fast-changing traffic environment and attendant human and social factors bring significant complexity to the VCS system and make it a typical cyber-physical-social system (CPSS), followed by the challenge of robust and efficient modeling of VCS systems. To cope with the complexity of social dimensions and optimize the decision-making process in the physical VCS, this article introduces the artificial societies, computational experiments, and the parallel execution (ACP) approach to the VCS system and develops a novel framework called parallel VCS (P-VCS). Three key components empower P-VCS to balance the physical environment, cyber networks, and human and social factors, namely, an artificial system that is used to parametrically describe the physical VCS, two types of computational experiments that simulate the decision process and evaluate different strategies, and the parallel execution mechanism that is used to characterize the system operation. To demonstrate the feasibility of the framework, we take participant selection under traffic events as an application example. Experimental results illustrate that the P-VCS-based parallel learning strategy maintains competitive performance in all cases.
TL;DR: In this paper , a parallel learning-based steering control method is proposed to solve the problem of data collection process for autonomous vehicles at high speeds by building a neural network based trajectory generative model (GeneratingNN) based on limited steering-trajectory raw data.
Abstract: Steering control for autonomous vehicles at high speeds is challenging due to the highly nonlinear vehicle dynamics. The traditional model-based controllers usually degrade significantly in this case. With the development of artificial intelligence, learning-based control methods are emerging as promising alternatives. These methods require a tremendous amount of training data to achieve acceptable performances. However, the data collection process is costly or inefficient. To solve this problem, we propose a parallel learning-based steering control method. Specifically, we first build a neural network (NN) based trajectory generative model (GeneratingNN) based on limited steering-trajectory raw data. The GeneratingNN can efficiently generate sufficient steering-trajectory data by enumerating the allowable steering actions sequences. Then, based on the raw data and generated data, we train another NN (RecallingNN) to learn the inverse mapping relationship between steering actions and trajectories. Hence, the RecallingNN can efficiently recall appropriate steering actions once given the previewed trajectory points. In addition, to further enhance the control accuracy and robustness, we use a simple feedback controller to handle the unmodeled dynamics and external disturbance. Testing results validate that the proposed method can achieve better tracking accuracy, stability and computational efficiency.
TL;DR: Wang et al. as mentioned in this paper proposed a fault-tolerant cooperative driving strategy for signal-free intersections by modeling potential vehicle failure types, aiming to keep a good balance between traffic safety and efficiency.
Abstract: Cooperative driving shows great potential to improve traffic safety and efficiency and has been well discussed in recent years. However, most existing studies focus on ideal traffic environments and ignore potential vehicle failures in traffic systems, which pose significant threats to traffic safety. Therefore, the fault-tolerant capacity of the existing cooperative driving strategies is questionable. To fill this research gap, this paper proposes a fault-tolerant cooperative driving strategy for signal-free intersections by modeling potential vehicle failure types, aiming to keep a good balance between traffic safety and efficiency. Notably, a rule-based fault-tolerant model is constructed to mitigate the threat of potential vehicle failures to traffic safety and efficiency, and to effectively recover the cooperative driving system after vehicle failures occur. Theoretical analysis and simulation results jointly demonstrate the promising performance of the proposed model in achieving fault tolerance and improving traffic efficiency.
TL;DR: In this paper , an approach to search for the Pareto optimal information flow topology off-line for the control of connected vehicles' platoon using a non-dominated sorting genetic algorithm was proposed.
Abstract: Information flow topology plays a crucial role in the control of connected autonomous vehicles. This paper proposes an approach to search for the Pareto optimal information flow topology off-line for the control of connected vehicles’ platoon using a non-dominated sorting genetic algorithm. Based on the obtained Pareto optimal information flow topology, the platoon's overall performance in terms of three main performance indices: tracking index, acceleration standard deviation, and fuel consumption, are all improved. Numerical simulations are used to validate the effectiveness of the proposed approach. In the simulation, the impact of different information flow topologies on the performance of the connected autonomous vehicles platoon is firstly investigated. The results show that more communication links can lead to better tracking ability. The smoothness of the velocity profile is consistent with fuel economy, while velocity profile's smoothness, fuel economy and communication efficiency are in contrary to the tracking index. Then, five cases are discussed using the Pareto optimal information flow topology. The results indicate that while ensuring the platoon's stability, the obtained Pareto optimal information flow topology can improve the tracking ability by 33.67% to 49.35%, and fuel economy by 7.181% to 16.93% and driving comfort up to 14.9%.
TL;DR: In this paper , an attention-based multi-modal model was proposed to improve decision-making performance and transparency of automated driving systems by providing multimodal explanations, especially when interacting with pedestrians.
Abstract: Automated driving desires better performance on tasks like motion planning and interacting with pedestrians in mixed-traffic environments. Deep learning algorithms can achieve high performance in these tasks with remarkable visual scene understanding and generalization abilities. However, when common scene-parsing methods are used to train end-to-end models, limitations of explainability in such algorithms inhibit their implementations in fully automated driving. The main challenges include algorithm performance deficiencies and inconsistencies, insufficient AI transparency, degraded user trust, and undermining human-AI interactions. This research aids the decision-making performance and transparency of automated driving systems by providing multi-modal explanations, especially when interacting with pedestrians. The proposed algorithm combines global visual features and interrelation features by parsing scene images as self-constructed graphs and using an attention-based module to capture the interrelationship among the ego-vehicle and other traffic-related objects. The output modules make decisions while simultaneously generating semantic text explanations. The results show that the fusion of the features from global frames and interrelational graphs improves decision-making and explanation predictions compared to two state-of-the-art benchmark algorithms. The interrelation module also enhances algorithm transparency by disclosing the visual attention used for decision-making. The importance of interrelation features on the two prediction tasks is further revealed along with the underlying mechanism of multitask learning on the datasets with hierarchical labels. The proposed model improves driving decision-making during pedestrian interactions with intelligible reasoning cues for building an appropriate mental model of automated driving performance for human users.
TL;DR: In this article , a Gaussian Process Regression (GPR) was used to predict the visual attention of a driver in a car environment, by tracking the position and orientation of the head.
Abstract: A smart vehicle should be able to understand human behavior and predict their actions to avoid hazardous situations. Specific traits in human behavior can be automatically predicted, which can help the vehicle make decisions, increasing safety. One of the most important aspects pertaining to the driving task is the driver's visual attention. Predicting the driver's visual attention can help a vehicle understand the awareness state of the driver, providing important contextual information. While estimating the exact gaze direction is difficult in the car environment, a coarse estimation of the visual attention can be obtained by tracking the position and orientation of the head. Since the relation between head pose and gaze direction is not one-to-one, this paper proposes a formulation based on probabilistic models to create salient regions describing the visual attention of the driver. The area of the predicted region is small when the model has high confidence on the prediction, which is directly learned from the data. We use Gaussian process regression (GPR) to implement the framework, comparing the performance with different regression formulations such as linear regression and neural network based methods. We evaluate these frameworks by studying the tradeoff between spatial resolution and accuracy of the probability map using naturalistic recordings collected with the UTDrive platform. We observe that the GPR method produces the best result creating accurate predictions with localized salient regions. For example, the 95% confidence region is defined by an area that covers 3.77% region of a sphere surrounding the driver.
TL;DR: The potential of ChatGPT for intelligent vehicle research from an IEEE TIV perspective is explored in this article , highlighting challenges and opportunities associated with these applications, as well as the feasibility of training intelligent vehicles using the same methods as chatGPT and reflecting the intelligence of intelligent vehicles in the context of human-machine shared control.
Abstract: This letter reports on a TIV DHW (decentralized and hybrid workshop) that explores the prospective influence of ChatGPT on research and development in intelligent vehicles. To assess the update capabilities of ChatGPT, we conducted tests involving both basic and technically relevant questions. Our preliminary testing revealed that ChatGPT's information can be updated and corrected at one time, but it may take some time for the changes to be reflected in ChatGPT's responses, so it may not always possess the latest knowledge regarding specific topics. We further discuss the prospective influence of ChatGPT on the field of intelligent vehicles, particularly possible applications of ChatGPT in areas like autonomous driving, human-vehicle interaction, and intelligent transportation systems, highlighting challenges and opportunities associated with these applications. Additionally, we address technical questions, such as the feasibility of training intelligent vehicles using the same methods as ChatGPT and the reflection of the intelligence of intelligent vehicles in the context of human-machine shared control. In conclusion, this letter presents a preliminary exploration of the potential of ChatGPT for intelligent vehicle research, from an IEEE TIV perspective, acknowledging the limitations and uncertainties of this emerging technology.
TL;DR: In this article , a hierarchical dynamic drifting controller (HDDC) is designed under both drifting maneuvers and typical cornering maneuvers to uniformly track the general path within and beyond stability limits.
Abstract: Taking full advantage of tire saturation in high sideslip drifting maneuvers can substantially improve the handling limits of autonomous vehicles. However, most singular motion control methods in conventional drifting controllers have been conducted, which cannot meet general driving needs. In this article, a hierarchical dynamic drifting controller (HDDC) is designed under both drifting maneuvers and typical cornering maneuvers to uniformly track the general path within and beyond stability limits. Specifically, the HDDC includes the path tracking layer, vehicle motion control layer, and actuator regulating layer. The first layer is designed to achieve accurate path tracking and generate the desired states, and it is algorithmically adaptable and confirmed by MPC and LQR. The second layer is proposed to integrate drifting and typical cornering control by the dynamic drifting inverse model. The third layer can achieve the steering system and wheel speed control to satisfy corresponding commands. Ultimately, the real-time performance of the controller is verified by the ECU hardware-in-the-loop platform. The results with MPC-HDDC and LQR-HDDC show that the HDDC can achieve integrated control of drifting and typical cornering maneuvers and possess high tracking accuracy.
TL;DR: Experimental results on an unmanned aerial vehicle (UAV) in an open semi-urban environment with multipath-free, line-of-sight (LOS) conditions are presented, showing that the developed framework achieves a 70.48 cm position root mean-squared error over a trajectory of 2.24 km.
Abstract: A framework that could achieve submeter-level-accurate horizontal navigation with carrier phase differential measurements from cellular signals is developed. This framework, termed CD-cellular, is composed of a base and a rover in a cellular environment, both making carrier phase measurements to the same cellular base transceiver stations (BTSs). The base shares its carrier phase measurements with the mobile rover, which in turn employs an extended Kalman filter to obtain a coarse estimate of its states, followed by a batch weighted nonlinear least squares (B-WNLS) estimator to solve for the integer ambiguities, and finally a point-solution WNLS to estimate its own states. The framework is designed to guarantee that after some time, the rover's position error remains below a pre-defined threshold with a desired probability. This is achieved by leveraging models of the BTS positions from stochastic geometry. Experimental results on an unmanned aerial vehicle (UAV) in an open semi-urban environment with multipath-free, line-of-sight (LOS) conditions are presented, showing that the developed framework achieves a 70.48 cm position root mean-squared error (RMSE) over a trajectory of 2.24 km, measured with respect to the UAV's navigation solution from its onboard GPS-inertial navigation system (INS).