2‐dimensional human‐like driver model for autonomous vehicles in mixed traffic

Abstract

Classical artificial potential approach of motion planning is extended for emulating human driving behaviour in two dimensions. Different stimulus parameters including type of ego‐vehicle, type of obstacles, relative velocity, relative acceleration, and lane offset are used. All the surrounding vehicles are considered to influence drivers' decisions. No emphasis is laid on vehicle control; instead, an ego vehicle is assumed to reach the desired state. The study is on human‐like driving behaviour modelling. The developed motion planning algorithm formulates repulsive and attractive potentials in a data‐driven way in contrast to the classical arbitrary formulation. Interaction between the stimulus parameters is explicitly considered by using multivariate cumulative distribution functions. Comparison of two‐dimensional (lateral and longitudinal) performance indicators with a baseline model and generative adversarial networks indicate the effectiveness and suitability of the developed motion planning algorithm in the mixed traffic environment.

Objectively Scoring the Human-Likeness of Artificial Driver Models

“…Road violation [7,11,25] Collision check [26] Dynamic Lateral Velocity [3] Longitudinal Velocity [3,[27][28][29][30] Lateral Acceleration [3,27,29] Longitudinal Acceleration [3,27,29,30] Longitudinal Jerk [29] Interaction Relative Velocity [3,27,29,30] Distance to the partner [3,29] Time-to-Collision (TTC) [28,29] Time Exposed Time-to-Collision (TET) [28] Max. Value for critical time gap when interacting [31] Post Encroachment Time (PET) [28] The parameters can be calculated for all samples of real and artificially generated behavioral data, provided that the spatiotemporal motion, road user classification, and information about the static environment, i.e., the map, are available.…”

“…The development of human-like driving capabilities in AVs is expected to enhance the ability of surrounding drivers to understand and anticipate the behavior of AVs, resulting in more natural interactions [2]. As a result, AVs are required to mimic human-like driving behavior [3]. Following, human-like behavior should be considered in the automated driving design, as mentioned by Hang et al [4] and in driver models noted by Lindorfer et al [5].…”

A novel merging strategy model considering the remaining distance in the acceleration lane

“…As the remaining distance in an acceleration lane decreases, the merging urgency of the merging vehicle increases. Therefore, it is necessary to investigate the impact of the remaining distance in the acceleration lane on the merging behaviour, explore the potential for safety improvements, build a more behaviourally sound merge strategy, and counter the effects of human driver inattention as a driving aid or for automated vehicles to merge in a human-like way [3][4][5][6][7].…”

Human-like Decision Making for Autonomous Vehicles at the Intersection Using Inverse Reinforcement Learning

“…Human beings have excellent scenario generalization, skill learning and emergency handling abilities. Autonomous vehicles (AVs) are publicly acceptable only if their driving behavior is comprehensible and comparable to that of human drivers [10][11][12]. Therefore, the study of the driving behavior of human drivers is an important aspect of intelligent driving vehicle research, towards making the automatic driving vehicles understand human driving mode, drive like humans and last enhancing people's acceptance of AVs.…”