Traffic Flow Management of Autonomous Vehicles Using Platooning and Collision Avoidance Strategies

Electronics, Journal Year: 2021, Volume and Issue: 10(10), P. 1221 - 1221

Published: May 20, 2021

Connected Autonomous Vehicles (AVs) promise innovative solutions for traffic flow management, especially congestion mitigation. Vehicle-to-Vehicle (V2V) communication depends on wireless technology where vehicles can communicate with each other about obstacles and make cooperative strategies to avoid these obstacles. Vehicle-to-Infrastructure (V2I) also helps use of infrastructural components navigate through different paths. This paper proposes an approach based swarm intelligence the formation evolution platoons maintain during collision avoidance practices using V2V V2I communications. In this paper, we present a two level improve AVs. At first level, reduce by forming study how platooning deal or in uncertain situations. We performed experiments challenging scenarios platoon’s evolution. second incorporate mechanism infrastructures. used SUMO, Omnet++ veins simulations. The results show significant improvement performance maintaining flow.

Language: Английский

---

An Empirical Study of DDPG and PPO-Based Reinforcement Learning Algorithms for Autonomous Driving

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 125094 - 125108

Published: Jan. 1, 2023

Autonomous vehicles mitigate road accidents and provide safe transportation with a smooth traffic flow. They are expected to greatly improve the quality of elderly or people impairments by improving their mobility due ease access transportation. sense driving environment navigate through it without human intervention. And, Deep Reinforcement Learning (DRL) is witnessed as powerful machine learning solution address sequential decision problem in autonomous vehicles. The detailed state-of-the-art work using DRL algorithms along future research directions discussed. Due high dimensional action space, two continuous space algorithms: Deterministic Policy Gradient (DDPG) Proximal Optimization (PPO) chosen complex problem. proposed DDPG PPO based decision-making models trained tested TORC simulator. Both for same number episodes lane keeping well multi-agent collision avoidance scenarios. To best our knowledge, this first paper present comparative performance analysis these algorithms, found perform better terms higher reward faster convergence than PPO. Hence, suitable option design model driving.

Language: Английский

---

Cross-domain cooperative route planning for edge computing-enabled multi-connected vehicles

Computers & Electrical Engineering, Journal Year: 2023, Volume and Issue: 108, P. 108668 - 108668

Published: March 15, 2023

Language: Английский

---

Traffic Speed Prediction Based on Time Classification in Combination With Spatial Graph Convolutional Network

IEEE Transactions on Intelligent Transportation Systems, Journal Year: 2022, Volume and Issue: 24(8), P. 8799 - 8808

Published: Aug. 26, 2022

With the advancement of automatic driving and smart city, it is critical to predict traffic information for management, planning, safety. When predicting information, spatial structure roads will also affect flow such as speed, occupancy rate, etc. The common method either merely focusing on temporal feature without considering structure, or extraction only applicable Euclidean which does not apply Non-Euclidean structure. This paper proposes a speed prediction based time classification in combination with Graph Convolutional Network. employs Gated Recurrent Unit extract correlation Network network’s In consideration varying features weekdays weekends dimension, divided into two types: weekends. Since road network change short term actual process, same graph convolution can reasonably be shared dimension after sections are fused training prediction. Finally, this proposed compared some baseline models prove performance. Generally speaking, strategy produces more accurate results PEMS_BAY METR_LA data sets than models.

Language: Английский

---

Multi-Level Objective Control of AVs at a Saturated Signalized Intersection with Multi-Agent Deep Reinforcement Learning Approach

Journal of Intelligent and Connected Vehicles, Journal Year: 2023, Volume and Issue: 6(4), P. 250 - 263

Published: Dec. 1, 2023

Reinforcement learning (RL) can free automated vehicles (AVs) from the car-following constraints and provide more possible explorations for mixed behavior. This study uses deep RL as AVs' longitudinal control designs a multi-level objectives framework trajectory decision-making based on multi-agent DRL. The saturated signalized intersection is taken research object to seek upper limit of traffic efficiency realize specific target control. simulation results demonstrate convergence proposed in complex scenarios. When prioritizing throughputs primary objective emissions secondary objective, both indicators exhibit linear growth pattern with increasing market penetration rate (MPR). Compared MPR 0%, be increased by 69.2% when 100%. adaptive cruise (LACC) under same MPR, also reduced up 78.8%. Under fixed throughputs, compared LACC, emission benefits grow nearly linearly increases, it reach 79.4% at 80% MPR. employs experimental analyze behavioral changes flow mechanism autonomy improve efficiency. method flexible serves valuable tool exploring studying behavior patterns autonomy.

Language: Английский

---