Rolling optimization decision-making method for unmanned vehicle at unprotected crossroad

Abstract

The invention provides a rolling optimization decision-making method for an unmanned vehicle at an unprotected crossroad. The method comprises the steps: acquiring the position and speed information of a vehicle and an obstacle, and the lane line information of the crossroad, and generating a reference track of the vehicle and a predicted surrounding vehicle track; then performing Frenet coordinate conversion on the vehicle information according to the reference trajectory of the vehicle so as to obtain the trajectory of the vehicle and the trajectory of surrounding vehicles under a Frenet coordinate system; then, establishing an unmanned vehicle crossroad decision model under a Frenet coordinate system to describe the movement of the vehicle at an unprotected crossroad; and finally, designing an unprotected crossroad decision-making controller based on model prediction control, solving an optimization problem, and enabling the longitudinal speed and the expected trajectory to act on a bottom layer controller, thereby realizing unprotected crossroad unmanned vehicle behavior decision-making based on model prediction control. According to the method, the relative position relation between the vehicle and surrounding objects is separated in the longitudinal direction and the transverse direction in the unmanned vehicle behavior decision-making process at the unprotected crossroad, and the collision danger between the vehicle and surrounding vehicles can be better described.

Description

technical field [0001] The invention belongs to the field of automobile driving control and relates to a rolling optimization decision-making method for an unmanned vehicle at an unprotected intersection. Background technique [0002] Unmanned driving is a current research hotspot, which is mainly composed of three parts: environmental perception, behavior decision-making, and motion control. Among them, behavior decision-making is the core of unmanned driving. The behavior decision-making system of unmanned vehicles is the key to determining its safety and stability, and it is the embodiment of its intelligence. The current decision-making methods used by unmanned vehicles are mainly divided into two categories: learning-based methods and rule-based methods. Learning-based decision-making methods require a large amount of data and long-term training, and their interpretability is poor; while rule-based methods are difficult to cover complex traffic scenarios such as unprot...

AI Technical Summary

Problems solved by technology

Learning-based decision-making methods require a large amount of data and long-term training, and their interpretability is poor; while rule-based methods are difficult to cover complex traffic scenarios such as unprotected intersections

Moreover, most of the existing decision-making models model the decision-making process as a Markov process. Due to its relatively high complexity and the need for transition probabilities between states, it is not conducive to the design of decision-making algorithms.

Images