An Online Evolving Framework for Advancing Reinforcement-Learning based Automated Vehicle Control

Vehicle automation is among the best possible solutions for traffic issues, including traffic accidents, traffic jams, and energy consumption. However, the user acceptance of automated vehicles is critical and is affected by riding comfort. In addition, human factors in automated vehicle control should be clear. This study evaluates the effect of different courses on driving comfort in automated vehicles using field experiments with 25 subjects. This study focused on lateral motion, but speed control was not targeted. Further, generating a path for obstacle avoidance and lane keeping, which have several constraining conditions, was also not targeted. Rendering a comfortable path is beneficial for developing an acceptable system as a car developer and for building new curves for automated or driving assistance systems from the perspective of construction. The automated vehicle drove at a speed of 30 km/h on four courses, namely, clothoid, two types of spline curves, and arc, based on the real intersection. Each participant sat on both the driver and passenger seat and answered a questionnaire. The experimental data indicated the clothoid course to be the most comfortable, while the arc was most uncomfortable for a significance level of 1%. These tendencies are applicable to driver and passenger seats, all genders, and experiences and will be beneficial for human factor research in automated vehicle control.

Download Full-text

Multiple Fault Diagnosis as Applied to Automated Vehicle Control

10.1115/imece1999-0083 ◽

1999 ◽

Author(s):

Adam S. Howell ◽

J. Karl Hedrick

Keyword(s):

Control System ◽

Diagnostic System ◽

Vehicle Control ◽

Longitudinal Control ◽

Multiple Faults ◽

Automated Vehicle ◽

Multiple Fault Diagnosis ◽

System Components ◽

Automated Vehicle Control ◽

Residual Processing

Abstract This paper addresses the problem of detecting multiple faults for the longitudinal control system of an automated vehicle. An existing fault diagnostic system which can isolate all single faults is extended to the diagnosis of multiple faults via improved residual processing in the form of fuzzy logic. The new diagnostic system is shown to correctly detect and isolate all single and multiple faults in a subset of the automated vehicle control system components.

Download Full-text

Deep Reinforcement Learning Based Left-Turn Connected and Automated Vehicle Control at Signalized Intersection in Vehicle-to-Infrastructure Environment

Information ◽

10.3390/info11020077 ◽

2020 ◽

Vol 11 (2) ◽

pp. 77 ◽

Cited By ~ 1

Author(s):

Juan Chen ◽

Zhengxuan Xue ◽

Daiqian Fan

Keyword(s):

Reinforcement Learning ◽

Control Method ◽

Signalized Intersection ◽

Signal Control ◽

Left Turn ◽

Automated Vehicle ◽

Whole Process ◽

Policy Gradient ◽

Experience Replay ◽

Automated Vehicle Control

In order to solve the problem of vehicle delay caused by stops at signalized intersections, a micro-control method of a left-turning connected and automated vehicle (CAV) based on an improved deep deterministic policy gradient (DDPG) is designed in this paper. In this paper, the micro-control of the whole process of a left-turn vehicle approaching, entering, and leaving a signalized intersection is considered. In addition, in order to solve the problems of low sampling efficiency and overestimation of the critic network of the DDPG algorithm, a positive and negative reward experience replay buffer sampling mechanism and multi-critic network structure are adopted in the DDPG algorithm in this paper. Finally, the effectiveness of the signal control method, six DDPG-based methods (DDPG, PNRERB-1C-DDPG, PNRERB-3C-DDPG, PNRERB-5C-DDPG, PNRERB-5CNG-DDPG, and PNRERB-7C-DDPG), and four DQN-based methods (DQN, Dueling DQN, Double DQN, and Prioritized Replay DQN) are verified under 0.2, 0.5, and 0.7 saturation degrees of left-turning vehicles at a signalized intersection within a VISSIM simulation environment. The results show that the proposed deep reinforcement learning method can get a number of stops benefits ranging from 5% to 94%, stop time benefits ranging from 1% to 99%, and delay benefits ranging from −17% to 93%, respectively compared with the traditional signal control method.

Download Full-text

Necessity of Automated Vehicle Control Customization: Experimental Results during Lane Changing

Transportation Research Record Journal of the Transportation Research Board ◽

10.1177/0361198118775871 ◽

2018 ◽

Vol 2672 (22) ◽

pp. 1-9

Author(s):

Naohisa Hashimoto ◽

Simon Thompson ◽

Shin Kato ◽

Ali Boyali ◽

Sadayuki Tsugawa

Keyword(s):

Vehicle Control ◽

Experimental Results ◽

Single Type ◽

Automated Driving ◽

Lane Changing ◽

Automated Vehicle ◽

Vehicle Trajectories ◽

Significant Difference ◽

The Subject ◽

Automated Vehicle Control

This study investigated the necessity of automated vehicle control customization for individual drivers via a lane-changing experiment involving 35 subjects and an automated minivan. The experiment consisted of two automated driving conditions: one in which the subject was unable to override vehicle controls, the other with the option to override when the subject felt it was necessary. The automated vehicle drove at a speed of 40 km/h along three kinds of planned paths for lane changing, generated by Bezier curves; the distance required for lane changing was varied to obtain the preferred path of each subject. Various data obtained during driving, including vehicle trajectories and steering angles produced by subjects were logged. After automated driving, a questionnaire was administered to each subject. The experimental data showed that there was a statistically significant difference between comfort when the vehicle drove along the subject’s preferred path, and when it drove along other paths. The results of the questionnaire indicated that 46% of the subjects preferred the planned path that most closely resembled their own. In addition, quantitative analysis of driving data found that approximately 69% of the subjects preferred an automated driving control that resembled their own. However, it was also observed that certain subjects were open to multiple types of automated vehicle control. The experimental results indicate that drivers will not necessarily accept a single type of automated vehicle control, therefore customization will be necessary to improve acceptance of automated driving.

Download Full-text