Research On Behavior Safety Strategy Of Unmanned Vehicles In Micro Traffic Environment

Unmanned vehicle can enhance road safety,reduce traffic flow and fuel consumption effectively.What's more,it is the important technology that can promote overall social development,which has great application value in urban transportation systems.The intersection as a typical complex traffic scene is an indispensable research object for the successful application of unmanned vehicles in urban roads.This paper,based on the support of the National R&D Projects “ Development and application of new multi-functional intelligent vehicle terminal(2018YFB1600701)”,mainly focus on the randomness of manual vehicles passing through the urban intersection,and studies the anti-collision motion control of unmanned vehicles.The main research contents are as follows:1.Establish a road traffic model at the intersection.In the PreScan,the road model is established with reference to the Xiaozhai intersection in Xi'an.In view of the complexity and uncertainty of the road traffic conditions,the actual traffic operation rules are considered and the collision situation is curtained to analyze the dynamic status between unmanned and manual vehicles.2.Predict the driving intention of the manual vehicle getting through the crossroad.According to he actual data collection,a continuous Hidden Markov Model for the manual vehicle driving intention at intersection is established.For the state characteristics of the vehicle object at intersection,select the Mixed Two-dimensional Gaussian Distribution to represent the probability relationship between observation sequence and the hidden state sequence.Modify the Baum-Welch algorithm on account of the distribution.The traffic scenarios are designed for different driving states.The Viterbi algorithm is used to verify the validity of the prediction model,with the introduction of the reference point and the sensitivity indicator.3.Predict the dynamic trajectory of the manual vehicle getting through the crossroad.Based on the premise that the driving intention is known,the trajectory of the manual vehicle is predicted to determine whether a collision will occur.The Extended Kalman Filter and the Particle Filter are used to predict the trajectory of the manual vehicle and the collision time.By optimizing the density function with the Unscented Kalman Filter,which has higheraccuracy,the Particle Filter is improved to suppress the degradation phenomenon.Collision scenarios are designed for three analyzed situations,and the accuracy of the two prediction models are verified and compared.4.Control the unmanned vehicle with tracking the re-planning anti-collision trajectory.Based on the principle of the Artificial Potential Field method,the trajectory planner of unmanned vehicle is established.In order to avoid the occurrence of the partial minimum point trap,the repulsion potential field function is improved and optimized.The polynomial fitting method is used to fit the reference trajectory of the trajectory planner output,for providing the tracking targets to the motion controller.Based on the Model Predictive Control algorithm,build the motion controller and verify its effectiveness by simulation.On the base of the driving intention and dynamic trajectory prediction of the manual vehicle,the motion control of unmanned vehicle at intersection is studied,which has a certain practical value.