Research On The Key Methods Of Tactical Planning For Intelligent Vehicle

Intelligent vehicle has become one of the most important development directions of automobile industry.Intelligent vehicle systems should complete the driving task safely,legally and efficiently as well as mature drivers.According to the road test report,intelligent vehicle can’t drive completely legally at present.The decision-making of drivers in driving is divided into three layers include strategic planning,tactical planning and motion planning.Strategic planning is to determine driving behaviors such as lane keeping,turning,U-turning,etc,while tactical planning is to determine which lane to drive into.The decision-making of these driving behaviors is the most critical decision to ensure full compliance with traffic rules and efficient driving.At present,the research of strategic planning is relatively mature,however the research of tactical planning on how to deal with traffic rules is not perfect.According to the research results,at present most intelligent vehicle decision-making systems deal with the traffic rules compliance and traffic participants avoidance in the motion planning.Its characteristic is that traffic rules and traffic participants are used as constraints to solve the optimal trajectory in motion planning.As a constraint condition,traffic rules have the problem of discontinuous constraint function,which leads to the difficulty of optimization.In view of the above problems,this paper studies the key methods of intelligent vehicle tactical planning based on the traffic rules compliance and traffic participants avoidance decoupling framework.The following aspects are mainly studied:First,a target lane decision-making method for intelligent vehicle is proposed.This paper summarizes the traffic rules,analyzes the constrains of the traffic rules on the vehicle driving process,and divides the target lane decision-making planning into the candidate Lane set generation module,the optimal Lane selection module,the expected driving lane and steering Light status determination module,referenced path planning module and desired speed planning module.Secondly,a motion prediction method based on driving intention recognition is proposed.In this paper,a driving intention recognition method is established.The recognition method isdivided into the driving behavior recognition in the road section area and the target lane recognition in the intersection area.The driving behavior model based on HMM and the target lane model based on Gaussian model are established.The prior probability model based on driving experience and traffic rules is established.The prior probability and likelihood probability calculated by intention model are used to calculate the driving intention probability.The driving intention with the highest probability is selected as the driving intention recognition result,and the motion prediction is carried out according to the driving intention recognition result and the traffic participants’ states.Build the intelligent vehicle simulation and verification platform,select typical scenarios to verify the target lane selection decision-making method established in this paper.The results show that the lane selection decision-making method established in this paper can ensure the intelligent vehicle to complete the driving task legally and efficiently.Ngsim data is used to verify the driving intention recognition method and motion prediction method.The results show that the driving intention recognition method established in this paper can recognize driving intention quickly,and the motion prediction method has higher accuracy than the kinematic model-based prediction approach.