Uncertainty-Aware Decision Making and Planning for ICV Based on Asymmetric Driving Aggressiveness

Precisely assessing driving threat of road segments could significantly enhance the driving efficiency of intelligent connected vehicles (ICV) within mixed traffic scenarios. Existing methods primarily concentrate on collision probabilities, resulting in an insufficient appraisal of asymmetrical hazard levels attributed to the various interactions. Meanwhile, the uncertainty and communication delay have great influence on ICV, and it is an issue that must be addressed when designing decision-making and planning model. Thus, this study proposes and formulates a new driving aggressiveness model after analyzing asymmetric interactions behaviors among vehicles with different types. Subsequently, aims to verify the capability of generating asymmetric interaction, the driving aggressiveness model is applied on lane-change decision-making and planning. Concretely, the aggressiveness-sensitive lanes-selection model is designed based on game theory, and the uncertainty-aware trajectory planning is developed by utilizing stochastic model predictive control (MPC) and the asymmetric driving aggressiveness. Finally, two naturalistic driving scenarios are utilized to verify the performance of the decision-making and planning model. The outcomes of simulations illustrate that the driving aggressiveness model introduces a novel perspective to assess the asymmetric driving threat. Meanwhile, the uncertainty-aware decision making and planning model can reduce the influence on interactive vehicles, and it has superior adaptability for dynamic and connected traffic environments.

I. Introduction

Distinct awareness of potential driving threat plays a pivotal role in influencing the embrace of autonomous vehicles by human drivers and passengers [1], [2]. Existing driving threat assessment methods often focus on the probability of collisions, overlooking the potential severity of collisions, which leads to a lack of asymmetrical evaluation for conflicting vehicles [3], [4]. The discrepancy from real-world driving conditions results in the generation of similar driving behaviors for different vehicles when decision-making models are grounded on symmetric threat model [5]. Besides, there have multiple uncertainties in intelligent connected vehicles, such as motion uncertainty, communication delay [6], [7]. The decision-making and planning without involving various uncertainties may lead to unsafe driving behavior or infeasible trajectory [8]. Hence, facilitating intelligent connected vehicle (ICV) to recognize asymmetric driving threats constitutes a crucial concern in uncertain driving environments. On the basis of our previous work [9], this study further explore the application of driving aggressiveness on decision-making and planning for ICV on uncertain driving environments.