A Survey of End-to-End Driving: Architectures and Training Methods

Tampuu, Ardi; Matiisen, Tambet; Semikin, Maksym; Fishman, Dmytro; Muhammad, Naveed

doi:10.1109/tnnls.2020.3043505

Cited by 153 publications

(83 citation statements)

References 89 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many metrics can be used to measure the behavior of agents [17]. For an autonomous driving system, safety and efficiency are the most concerned performance metrics.…”

Section: Performance Evaluation Metricsmentioning

confidence: 99%

Harmonious Lane Changing via Deep Reinforcement Learning

Wang

2022

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

The development of autonomous driving has attracted extensive attention in recent years, and it is essential to evaluate the performance of autonomous driving. However, testing on the road is expensive and inefficient. Virtual testing is the primary way to validate and verify self-driving cars, and the basis of virtual testing is to build simulation scenarios. In this paper, we propose a training, testing, and evaluation pipeline for the lane-changing task from the perspective of deep reinforcement learning. First, we design lane change scenarios for training and testing, where the test scenarios include stochastic and deterministic parts. Then, we deploy a set of benchmarks consisting of learning and non-learning approaches. We train several state-of-the-art deep reinforcement learning methods in the designed training scenarios and provide the benchmark metrics evaluation results of the trained models in the test scenarios. The designed lane-changing scenarios and benchmarks are both opened to provide a consistent experimental environment for the lane-changing task 1 .

show abstract

“…Many metrics can be used to measure the behavior of agents [17]. For an autonomous driving system, safety and efficiency are the most concerned performance metrics.…”

Section: Performance Evaluation Metricsmentioning

confidence: 99%

Harmonious Lane Changing via Deep Reinforcement Learning

Wang

2022

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

show abstract

“…Evaluation of dynamic performance is important since the algorithm implements the logic of autonomous vehicles in motion and in a holistic way [41]. It is also an important step towards safety measures and interpretability of end-to-end driving systems [1,2,3,4]. We propose such an integrated algorithm consisting of a new UNet architecture for multi-task learning, a path prediction model [23] using UNet's output, a modified lateral controller, and a modeland learning-based longitudinal controller.…”

Section: Contributionmentioning

confidence: 99%

“…End-to-end driving system with a single deep neural network (DNN) is an emerging technology in autonomous vehicles [1,2,3,4]. The system is a pipeline consisting of perception sensors, DNN, and control actuators [1,2,3] with a data flow from sensors to DNN to path planning to controllers to actuators for making driving decisions of steering, acceleration, or braking in an end-to-end, autonomous, and real-time manner [1,2,3,5]. Since Pomerleau's pioneering work in the 1980s [6], a variety of end-to-end DNNs have been proposed for various tasks in autonomous driving [5,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24].…”

Section: Introductionmentioning

confidence: 99%

End-to-End Multi-Task Deep Learning and Model Based Control Algorithm for Autonomous Driving

Lee¹,

Liu²

2021

Preprint

View full text Add to dashboard Cite

End-to-end driving with a deep learning neural network (DNN) has become a rapidly growing paradigm of autonomous driving in industry and academia. Yet safety measures and interpretability still pose challenges to this paradigm. We propose an end-toend driving algorithm that integrates multi-task DNN, path prediction, and control models in a pipeline of data flow from sensory devices through these models to driving decisions. It provides quantitative measures to evaluate the holistic, dynamic, and real-time performance of end-to-end driving systems, and thus allows to quantify their safety and interpretability. The DNN is a modified UNet, a well known encoder-decoder neural network of semantic segmentation. It consists of one segmentation, one regression, and two classification tasks for lane segmentation, path prediction, and vehicle controls. We present three variants of the modified UNet architecture having different complexities, compare them on different tasks in four static measures for both single and multi-task (MT) architectures, and then identify the best one by two additional dynamic measures in real-time simulation. We also propose a learning-and model-based longitudinal controller using model predictive control method. With the Stanley lateral controller, our results show that MTUNet outperforms an earlier modified UNet in terms of curvature and lateral offset estimation on curvy roads at normal speed, which has been tested in a real car driving on real roads.

show abstract

“…This task is the output of all the previous tasks and based on the all-stacked information the AV make decision and control steering, acceleration, and deceleration. We refer readers to [17] as a comprehensive study about end-to-end autonomous driving.…”

Section: Autonomous Drivingmentioning

confidence: 99%

“…The emergency vehicle (emg) always starts travel from ( = 0) in the emergency lane ( ) and has a random departure speed of [20,50] (m/s). The ego vehicle will be departure with a random position in the first 200 meters of the highway in front of the emg vehicle in one of three lanes.…”

Section: Environmentmentioning

confidence: 99%

Decision-making of an autonomous vehicle in the presence of emergency vehicle using deep reinforcement learning

Shoaraee¹

View full text Add to dashboard Cite

Autonomous Vehicles are the future of road transportation where they can increase safety, efficiency, and productivity. In this thesis, we address a new edge case in autonomous driving when one autonomous vehicle is approached by an emergency vehicle and needs to make the best decision. To achieve the desired behavior and learn the sequence decision process, we trained our autonomous vehicle with the help of Deep Reinforcement Learning algorithms and compared the results with rule-based algorithms. The driving environment for this study was developed by using Simulation Urban Mobility as an open-source traffic simulator. The proposed solution based on Deep Reinforcement Learning has a better performance compared to the rule-based solution as a baseline both in normal driving situations and when an emergency vehicle is approaching.

show abstract

A Survey of End-to-End Driving: Architectures and Training Methods

Cited by 153 publications

References 89 publications

Harmonious Lane Changing via Deep Reinforcement Learning

Harmonious Lane Changing via Deep Reinforcement Learning

End-to-End Multi-Task Deep Learning and Model Based Control Algorithm for Autonomous Driving

Decision-making of an autonomous vehicle in the presence of emergency vehicle using deep reinforcement learning

Contact Info

Product

Resources

About