A Hybrid Deep Reinforcement Learning and Optimal Control Architecture for Autonomous Highway Driving DOI Creative Commons
Nicola Albarella, Dario Giuseppe Lui, Alberto Petrillo

et al.

Energies, Journal Year: 2023, Volume and Issue: 16(8), P. 3490 - 3490

Published: April 17, 2023

Autonomous vehicles in highway driving scenarios are expected to become a reality the next few years. Decision-making and motion planning algorithms, which allow autonomous predict tackle unpredictable road traffic situations, play crucial role. Indeed, finding optimal decision all different is challenging task due large complex variability of scenarios. In this context, aim work design an effective hybrid two-layer path architecture that, by exploiting powerful tools offered emerging Deep Reinforcement Learning (DRL) combination with model-based approaches, lets properly behave conditions and, accordingly, determine lateral longitudinal control commands. Specifically, DRL-based high-level planner responsible for training vehicle choose tactical behaviors according surrounding environment, while low-level converts these choices into actions be imposed through optimization problem based on Nonlinear Model Predictive Control (NMPC) approach, thus enforcing continuous constraints. The effectiveness proposed hierarchical hence evaluated via integrated vehicular platform that combines MATLAB environment SUMO (Simulation Urban MObility) simulator. exhaustive simulation analysis, carried out non-trivial scenarios, confirms capability strategy

Language: Английский

The impacts of connected autonomous vehicles on mixed traffic flow: A comprehensive review DOI
Yuchen Pan, Yu Wu, Lu Xu

et al.

Physica A Statistical Mechanics and its Applications, Journal Year: 2024, Volume and Issue: 635, P. 129454 - 129454

Published: Jan. 3, 2024

Language: Английский

Citations

23
Quantitative Identification of Driver Distraction: A Weakly Supervised Contrastive Learning Approach DOI
Haohan Yang, Haochen Liu, Zhongxu Hu

et al.

IEEE Transactions on Intelligent Transportation Systems, Journal Year: 2023, Volume and Issue: 25(2), P. 2034 - 2045

Published: Sept. 27, 2023

Accurate recognition of driver distraction is significant for the design human-machine cooperation driving systems. Existing studies mainly focus on classifying varied distracted behaviors, which depend heavily scale and quality datasets only detect discrete categories. Therefore, most data-driven approaches have limited capability recognizing unseen activities cannot provide a reasonable solution downstream applications. To address these challenges, this paper develops vision Transformer-enabled weakly supervised contrastive (W-SupCon) learning framework, in behaviors are quantified by calculating their distances from normal representation set. The Gaussian mixed model (GMM) employed clustering, centralizes distribution set to better identify behaviors. A novel behavior dataset other three ones evaluation, experimental results demonstrate that our proposed approach has more accurate robust performance than existing methods unknown activities. Furthermore, rationality levels different evaluated through skeleton poses. constructed demo videos available at https://yanghh.io/Driver-Distraction-Quantification .

Language: Английский

Citations

28
Adaptive eco-cruising control for connected electric vehicles considering a dynamic preceding vehicle DOI
Yichen Liang, Haoxuan Dong, Dongjun Li

et al.

eTransportation, Journal Year: 2023, Volume and Issue: 19, P. 100299 - 100299

Published: Dec. 3, 2023

Language: Английский

Citations

24
An in-depth evaluation of deep learning-enabled adaptive approaches for detecting obstacles using sensor-fused data in autonomous vehicles DOI
Abhishek Thakur, Sudhanshu Mishra

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108550 - 108550

Published: May 9, 2024

Language: Английский

Citations

14
Estimation of Lateral Velocity and Cornering Stiffness in Vehicle Dynamics Based on Multi-Source Information Fusion DOI
Guoying Chen,

Jun Yao,

Zhenhai Gao

et al.

SAE International journal of vehicle dynamics, stability, and NVH, Journal Year: 2024, Volume and Issue: 8(1)

Published: Jan. 4, 2024

<div>To address the challenge of directly measuring essential dynamic parameters vehicles, this article introduces a multi-source information fusion estimation method. Using intelligent front camera (IFC) sensor to analyze lane line polynomial and kinematic model, vehicle’s lateral velocity sideslip angle can be determined without extra expenses. After evaluating strengths weaknesses two aforementioned techniques, approach for is proposed. This extracts characteristics calculate allocation coefficient. Subsequently, outcomes from techniques are merged, ensuring rapid convergence under steady-state conditions precise tracking in scenarios. In addition, we introduce tire parameter online adaptive module (TPOAM) continually update such as cornering stiffnesses, with its effectiveness demonstrated through DLC slalom simulation tests. dual extended Kalman filter (DEKF) observer, allows joint vehicle states parameters. Ultimately, offer cost-effective method vital support motion control autonomous driving.</div>

Language: Английский

Citations

13
Coordinated control of path tracking and yaw stability for distributed drive electric vehicle based on AMPC and DYC DOI
Dongmei Wu, Yuying Guan, Xin Xia

et al.

Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 26, 2024

Maintaining both path-tracking accuracy and yaw stability of distributed drive electric vehicles (DDEVs) under various driving conditions presents a significant challenge in the field vehicle control. To address this limitation, coordinated control strategy that integrates adaptive model predictive (AMPC) direct moment (DYC) is proposed for DDEVs. The strategy, inspired by hierarchical framework, upper layer lower Based on linear time-varying (LTV MPC) algorithm, effects prediction horizon weight coefficients are compared analyzed first. According to aforementioned analysis, an AMPC controller with variable designed considering change speed layer. involves DYC based quadratic regulator (LQR) technique. Specifically, intervention rule determined threshold rate error phase diagram sideslip angle. Extensive simulation experiments conducted evaluate different conditions. results show that, low adhesion conditions, have been improved 21.58% 14.43%, respectively, AMPC. Similarly, high 44.30% 14.25%, coordination LTV MPC DYC. indicate effective across speeds. Furthermore, successfully enhances while maintaining good even extreme

Language: Английский

Citations

7
Risk-aware lane-change trajectory planning with rollover prevention for autonomous light trucks on curved roads DOI
Hefeng Zhan, Gang Wang, Xin Shan

et al.

Mechanical Systems and Signal Processing, Journal Year: 2024, Volume and Issue: 211, P. 111126 - 111126

Published: Feb. 6, 2024

Language: Английский

Citations

7
FedBEVT: Federated Learning Bird's Eye View Perception Transformer in Road Traffic Systems DOI Creative Commons
Rui Song, Runsheng Xu, Andreas Festag

et al.

IEEE Transactions on Intelligent Vehicles, Journal Year: 2023, Volume and Issue: 9(1), P. 958 - 969

Published: Aug. 31, 2023

Bird's eye view (BEV) perception is becoming increasingly important in the field of autonomous driving. It uses multi-view camera data to learn a transformer model that directly projects road environment onto BEV perspective. However, training often requires large amount data, and as for traffic are private, they typically not shared. Federated learning offers solution enables clients collaborate train models without exchanging but parameters. In this paper, we introduce FedBEVT, federated approach perception. order address two common heterogeneity issues FedBEVT: (i) diverse sensor poses, (ii) varying numbers systems, propose approaches - Learning with Camera-Attentive Personalization (FedCaP) Adaptive Multi-Camera Masking (AMCM), respectively. To evaluate our method real-world settings, create dataset consisting four typical use cases. Our findings suggest FedBEVT outperforms baseline all cases, demonstrating potential improving

Language: Английский

Citations

14
Lateral control for autonomous vehicles: A comparative evaluation DOI Creative Commons
Antonio Artuñedo, Marcos Moreno-Gonzalez, Jorge Villagrá

et al.

Annual Reviews in Control, Journal Year: 2023, Volume and Issue: 57, P. 100910 - 100910

Published: Nov. 3, 2023

The selection of an appropriate control strategy is essential for ensuring safe operation in autonomous driving. While numerous strategies have been developed specific driving scenarios, a comprehensive comparative assessment their performance using the same tuning methodology lacking literature. This paper addresses this gap by presenting systematic evaluation state-of-the-art model-free and model-based strategies. objective to evaluate contrast these controllers across wide range reflecting diverse needs vehicles. To facilitate analysis, set metrics selected, encompassing accuracy, robustness, comfort. contributions research include design methodology, use two novel stability comfort comparisons through extensive simulations real tests experimental instrumented vehicle over trajectories.

Language: Английский

Citations

12
Secure Cooperative Localization for Connected Automated Vehicles Based on Consensus DOI Creative Commons
Xin Xia, Runsheng Xu, Jiaqi Ma

et al.

IEEE Sensors Journal, Journal Year: 2023, Volume and Issue: 23(20), P. 25061 - 25074

Published: Sept. 11, 2023

In this paper, we present secure cooperative localization for connected automated vehicles (CAVs) based on consensus estimation through leveraging shared but possibly attacked sensory information from multiple adjacent vehicles. First, the communication topology between CAVs, node kinematic model, and measurement model each vehicle are introduced. Then, a Kalman filter (CKIF) is applied to fuse Since might be attacked, an attack detection algorithm general likelihood ratio test (GLRT) adopted. A delay-prediction framework proposed maintain accuracy real-time performance of algorithm. Next, rule-based isolation method used defend attack. Finally, validated in extensive numerical simulation experiments. The results confirm that manner leads better resilience under attacks.

Language: Английский

Citations

11