Research On Traffic Conflict Characteristics And Micro Behaviour Model At Urban Non-motorized Shared Space

The non-motorized shared space,where pedestrians,bicycles and e-bikes share the right-of-way,is one of the important forms of urban non-motorized transportation.The non-motorized shared spaces save the resource of urban transportation land and provide a flexible and free travel environment for vulnerable road users.However,the safety issue of non-motorized shared space should not be ignored,because all kinds of road users with different speeds are allowed to travel in such unsupervised space.which leads to the frequent occurrence of traffic conflicts,and even the traffic accidents between various types of shared-space users.Pedestrians are also often exposed to higher risks due to these conflicts.The non-motorized traffic safety issue has not paid enough attention in the process of urbanization dominated by motor vehicles.However,with the deepening of the concept of people-oriented transportation,creating safe and comfortable non-motorized spaces have become an urgent requirement for urban development."Improving the quality of walking,bicycling and other nonmotorized types" is also clearly stated in the "Outline for the Construction of a Strong Transportation Country".However,safety theories and techniques for non-motorized shared spaces are scanty.Traffic conflict technique developed with a focus on motor vehicle is also difficult to be directly applied on the nonmotorized shared space users with high movement flexibility and complex trajectories.In view of this,this study carried out a systematic study of the traffic conflicts at non-motorized shared spaces.The traffic conflict techniques at non-motorized shared spaces were studied,conflict characteristics and conflict behaviours were analyzed,and conflict behaviour prediction models were constructed.Firstly,the definition of traffic conflict at non-motorized shared spaces was defined,and the criteria for identifying and grading the traffic conflict severity were proposed,based on the motor vehicle traffic conflict techniques and according to the traffic characteristics of non-motorized shared space.The conflict characteristic indicators were constructed and the calculating method is given.Based on the videos of 12 non-motorized shared space locations in Futian District,Shenzhen city,2386 traffic conflicts were identified using the nonmotorized shared space traffic conflict technique proposed in this paper.1052 traffic conflict trajectories were extracted,and 137460 trajectory data points were obtained by time-frames.A multidimensional database including traffic volume,traffic conflicts,traffic conflict trajectories,and traffic accidents was built.Based on the identified traffic conflict data,the traffic conflict characteristics of the non-motorized shared space were studied.The severity,conflict types,participant characteristics and time distributions of the traffic conflicts at non-motorized shared space were analyzed.The relationships between the number of conflicts and traffic volume,conflict rat io and traffic ratio were compared,and it was found that the traffic complexity and traffic volume are the key factors affecting the conflicts at non-motorized shared space.Based on the conflict influencing factors,the conflict coefficients that can be used to describe the conflict distribution and assess the safety condition at nonmotorized shared space were proposed.The characteristics of evasive actions with different severity levels and different characteristics of traffic participants were analyzed.The key avoiding behaviours of "swerve" were identified,and the distributions of swerve directions were studied.The multinomial logit model and stochastic parametric logit model were constructed.13 significant factors influencing the severity of traffic conflicts at non-motorized shared spaces under heterogeneous conditions and the random parameter variable "e-bikes " were identified.Then,based on the cognitive model of driving behavio urs,a definition and identification method of key points of traffic conflict behaviour integrating microscopic behavioural and cognitive changes are proposed.Conflict behaviour phases were delineated and a conflict behaviour framework was constructed.Conflict behaviour indicators covering distance,area,time and cognition were proposed,and the calculation method of indicators was given.Based on the proposed conflict behaviour indicators,the microscopic behaviours of conflicts in opposite-direction and same-direction at nonmotorized shared spaces were analyzed.The conflict behaviours were compared under different conflict severity,types of road users,gender and whether they carried large objects.The mechanisms of conflict occurrence were analyzed,the conflict behaviour patterns were revealed,and a traffic con flict behaviourcognition model at non-motorized shared space was constructed.Then,Lotka–Volterra model was used to prove that the conflict between pedestrian and e-bikes is the most important conflict form at non-motorized shared space,and the conflict behaviour of this conflict type was predicted.A multi-agent model for predicting the conflict behaviours of pedestrians and ebikes was constructed.The two key steps of reward function estimation based on inverse reinforcement learning and policy functi on estimation based on MACK algorithm were emphasized.This model introduces the Logistic Stochastic Optimal Response Equilibrium,and the main advantage is that it considers the sub-optimal behaviour of agents(conflict participants).Finally,the pedestrian and e-bikes conflict behaviour prediction model was proposed in this paper,which was applied to predict the conflict behaviour and validated by using the data of specific non-motorized shared spaces in Shenzhen.Highly nonlinear reward functions were obtained for pedestrians and e-bikes,respectively.The behavioural preferences related to the distance,swerve and speed were dissected,competitive and cooperative relationships of conflicting behaviours were explored.The effectiveness and accuracy of the conflict behaviour prediction model were also demonstrated visually by comparing the predicted trajectory with the actual trajectory.A set of definition,identification and severity discrimination methods applicable to non-motorized shared space traffic conflicts were proposed.The traffic conflict characteristics at non-motorized shared space were analyzed,and the behaviour-cognition model of traffic conflicts at non-motorized shared space were constructed.Based on the behavioural game theory and inverse reinforcement learning theory,a multi-agent pedestrian and e-bikes conflict behaviour prediction model was established and verified.The research results help to enrich the theoretical system of traffi c conflict technology,provide theoretical support for improving the severe safety posture of urban slowmoving traffic participants,and provide decision basis for s low-moving space in engineering practice.