Studies Of Dynamics And Intelligent Development In Transportation System

Transportation system is an important component in modern society.It is necessary to understand different traffic phenomena in order to achieve transportation efficiency,and potential economic and social benefits.With the ever-increasing traffic all over the world,effective control and optimization of transportation system has always been demanded.This thesis investigates comprehensive traffic system dynamics and proposes novel optimization algorithms for development of intelligent transportation system.Common traffic systems are studied in this thesis,including urban street,highway,intersection,urban network and intelligent vehicle.Urban street consists of a wide variety of traffic participants including cars,bicycles,pedestrians and so forth.Significant influence of interactions between different entities on traffic system has been observed.In this thesis,we focus on the interaction between vehicles and pedestrians,and its effect on different traffic performance including transportation efficiency,traffic safety and energy consumption.Pedestrian noncompliance crossings are investigated,as well as vehicle lane-changing behaviors to avoid pedestrians.While vehicle longitudinal behaviors have been studied extensively in the literature,lane-changing movements receive much less attention,especially in highway system with bottlenecks.Because of traffic accident,some lanes are closed along the highway,and the bottleneck due to lane reduction is formed.Vehicles have to change lane to cross the forbidden area,that leads to complex interactions among vehicles before the incident place.Due to malfunction,some vehicle travels much slower than the other ones,that results in a moving bottleneck.The models of traffic bottlenecks are formulated in this thesis,and the influence of the bottleneck on traffic performance is studied.As crucial nodes in the city traffic network,intersections have significant impact on the traffic performance.Effective regulation of vehicle flow at intersections has always been an important issue in traffic systems.The intersection system is extensively studied in this thesis.The influence of vehicle flow,pedestrian flow,speed limit,driver rational rate and so forth on the intersection performance is investigated.A multi-objective optimization algorithm is proposed to achieve better traffic performance including reducing traffic delays,enhancing safety and achieving energy economy.Signal setting and lane assignment are optimized simultaneously.We find the performance improvement by the combined optimization method can not be achieved by the signal optimization alone.Furthermore,multiple turning lanes are often used at large intersections to discharge strong turning traffic flows.While transportation efficiency can be usually improved by multiple turning lanes,traffic safety tends to be compromised.The effects of multiple turning lanes on large intersection systems are investigated.A multi-objective optimization algorithm is proposed to design optimal lane assignment at large intersections.It is observed that turning lane number has significant influence on large intersection performance,and improvements in the optimization results can be obtained if turning lane assignment is incorporated in the design space.Traffic network congestion in urban area has been a serious problem all over the world.Transportation system suffers from long delays derived from intersections in network.This thesis proposes two coordinated signal optimization methods to improve network performance.Since popular network models in the literature are often over simplified,we develop a traffic network model considering many realistic issues such as route choices and vehicle turning behaviors.A static signal optimization method is proposed.Signal timing,left-turn signal type and lane assignment at each intersection in the network are optimized based on the prior information of the network such as traffic demands.It is an offline algorithm,that can handle large networks with many intersections.The other optimization method is dynamic predictive network signal control.It is an online algorithm that considers instant traffic environment information.The optimal Pareto solution set can be obtained at each control step by predicting the system behavior in the prediction horizon with different designs.The dynamic optimization method is robust to uncertainty,disturbances and model mismatch.The online computational load can be significantly reduced to be acceptable with the help of CPU parallel computing.Autonomous vehicle intelligence has great potential to improve mobility,enhance traffic safety and achieve energy economy in intelligent transportation system.However,the state-of-the-art autonomous driving technologies such as sensing,decision making and motion planning have not achieved the same performance as those of good human drivers.In this thesis,we present a study on the autonomous ground vehicle planning and control.A dynamic high-level motion and trajectory planning method is proposed.Following the idea of defensive driving,the autonomous vehicle is supposed to avoid careless-driving and aggressive neighbors.The trajectory is optimized based on the desired motion plan considering stability,safety,travel efficiency,driving comfort and so forth.Moreover,an optimization on the low-level adaptive cruise control is carried out.We propose a PID form control in the car-following mode.A wide variety of scenarios are created to test the effectiveness of the proposed method.