Research On Multi-Vehicle Cooperative Merging Strategy Based On Self-Driving Vehicles

Cooperation is highly needed to ensure safety and efficiency when vehicles are forced to change lane,which is normally caused by road structure changes or obstacles.Cooperative merging strategy of intelligent vehicles is proposed in this paper and a semi-physical simulation platform based on micro-IV is designed and build in order to validate cooperative driving strategy.In this paper,the merging problem is divided into random traffic merging and platoon merging,according to different traffic organization form.As to random traffic with less constrains between vehicles,a traffic merging strategy aiming to improve throughput was proposed in this paper.According to different traffic parameters,large selections of typical traffic merging scenarios were selected for research.For vehicles in different scenarios,decision sequence was encoded and optimal merging decision was obtained by Genetic Algorithm based on RSSR.Those optimal decisions were used to train Classification and Regression Tree.Specifically,the environmental feature was described by vehicle state and relationship between other vehicles around.Then relationship between environmental features and decision was described by Classification and Regression Tree.Compared with previous merging strategy by simulation,it was showed that merging strategy based on CART could effectively mitigate disturbance on traffic flow,brought by merging maneuver and maintain high through efficiency even in large flow circumstance.Moreover,this method is also rather robust to environmental perception errors,such as positioning error,which may exist in implementation.As to platoon which has stronger constrains between vehicles,Intra-platoon centralized control was applied and performance index of merging trajectory was formulated.Model Predictive Control algorithm was adopted to plan a merging trajectory,which is both merging time optimal and riding comfortable.This control method keeps optimizing the control input according to performance index in a certain range of time domain.Moreover,with the help of feedback and prediction state correction,this method is able to respond to the real-time traffic situation and optimize the path dynamically.Finally,a semi-physical simulation platform based on micro-IV and motion capture device is build in order to validate cooperative driving strategy.Control loop of the system is as follows:The vehicle is captured by the infrared camera and its pose is analyzed and transmitted to the host computer.Then host computer calculates the control instruction according to scheduling program and broadcasts to all the micro-? via Xbee Pro,the miniature car parses the control instructions and carries out the operation.Interactive Multi-Model filter was also adopted to improve the stability of positioning.