Deep Learning Model Predictive Control for Autonomous Driving in Unknown Environments

Rheem

Journal of Cognitive Engineering and Decision Making

et al. 2022

Advances in automated driving systems (ADSs) have shifted the primary responsibility of controlling a vehicle from human drivers to automation. Framing driving a highly automated vehicle as teamwork can reveal practical requirements and design considerations to support the dynamic driver–ADS relationship. However, human–automation teaming is a relatively new concept in ADS research and requires further exploration. We conducted two literature reviews to identify concepts related to teaming and to define the driver–ADS relationship, requirements, and design considerations. The first literature review identified coordination, cooperation, and collaboration (3Cs) as core concepts to define driver–ADS teaming. Based on these findings, we propose the panarchy framework of 3Cs to understand drivers’ roles and relationships with automation in driver–ADS teaming. The second literature review identified main challenges for designing driver–ADS teams. The challenges include supporting mutual communication, enhancing observability and directability, developing a responsive ADS, and identifying and supporting the interdependent relationship between the driver and ADS. This study suggests that the teaming concept can promote a better understanding of the driver–ADS team where the driver and automation require interplay. Eventually, the driver–ADS teaming frame will lead to adequate expectations and mental models of partially automated vehicles.

Section: Methodsmentioning

confidence: 99%

Teaming with Your Car: Redefining the Driver–Automation Relationship in Highly Automated Vehicles

Rheem

Journal of Cognitive Engineering and Decision Making

et al. 2022

“…Actor-critic (AC) method, which uses the approximate value function to guide the process of policy updating, greatly realizes the convergence of finding the optimal process. Critics approximate method is used to update action value function Q PI (s, a), action network using the method of random strategy under the guidance of commenting on the network, the stochastic gradient PI (a | s theta) update [18].…”

Section: Deep Reinforcement Learning Methodsmentioning

confidence: 99%

Deep Learning in Autonomous Driving

2020

Autonomous driverless technology, as an emerging technology of cross-integration of multiple fields, can control vehicles through autonomous intelligent control strategies, bring more safe and reliable driving environment and efficient driving scheme, which has very important research significance for improving the safety and efficiency of real road traffic. This paper mainly studies the application of deep learning in autonomous driving. This paper first analyzes the deep reinforcement learning method, and proposes an ODDPG autonomous driving decision method based on supervised training feature network for the real complex road autonomous driving decision problem. The feature extraction network is obtained in advance from imitation learning to improve the feature sample collection efficiency of reinforcement learning. ODDPG algorithm shows faster learning speed and better reward distribution of interactive data in the training process.

“…Therefore, when analyzing a scene, DL is reasonably insensitive to variations of environmental conditions, yet requires a large amount of high-quality data to achieve high accuracy [6]. Neural networks can learn complex interactions between features, which is beneficial for autonomous driving in dynamic environments [7]. Steering a car through traffic constitutes a complex task that is very hard to cast into algorithms.…”

Section: Introductionmentioning

confidence: 99%

“…Many studies have applied DL in predicting the steering angle of AVs [2], [18], [19], [7], [20]. Kuutti, Bowden [4] and Oussama and Mohamed [21] surveyed DL applications in AV control.…”

Section: Introductionmentioning

confidence: 99%

A Survey on Deep Learning for Steering Angle Prediction in Autonomous Vehicles

et al. 2020

Steering angle prediction is critical in the control of Autonomous Vehicles (AVs) and has attracted the attention of researchers, manufacturers, and insurance companies in the automotive industry. Different Deep Learning (DL) architectures have been applied to predict the steering angle of AVs in various scenarios. A survey on steering angle prediction based on deep learning algorithms can help expert researchers identify those areas that require development. Also, novice researchers can use the survey as a starting point. In this paper, we present a broad study on the recent advances made in DL architectures that covers the steering angle prediction of AVs. A new comprehensive taxonomy of the application of DL in steering angle prediction of AVs is created. The survey presents a concise research summary synthesis, and analysis. It is found that most researchers depend on Convolutional Neural Network (CNN) over other DL architectures in predicting the steering angle of autonomous driving vehicles. Also identified are open research problems. The prominent challenge facing DL-based steering angle prediction of AVs is lack of sufficient real-world datasets, which means that researchers largely depend on data generated from simulated environments. Lastly, alternative viewpoints to solve the identified open research challenges are proposed, pointing towards promising future research directions.