Lu Shoufeng

Q-Learning for Adaptive Traffic Signal Control Based on Delay Minimization Strategy

The goal of the paper is to test the performance of Q-learning for adaptive traffic signal control. For Q-learning algorithm, the state is total delay of the intersection, and the action is phase green time change. The relationship between phase green time change and action space is discussed. The performance between Q-learning and fixed cycle signal setting for isolated intersection is compared. The computation results show that Q-learning for traffic signal control can achieve lesser delay for variable traffic condition.Incremental multistep Q learning (Q( lambda )) combines Q learning and TD(lambda ). Theoretically, Q(lambda ) has better performance than Q learning. The goal of the paper is to test the performance of Q(lambda ) for adaptive traffic signal control. For Q(lambda ), the state is total delay of the intersection, and the action is phase green time change. The relationship between phase green time change and action space is discussed. The performance between Q(lambda) learning and fixed cycle signal setting for isolated intersection is compared. The computation results show that Q(lambda ) learning for traffic signal control can achieve lesser delay for variable traffic condition.

Based on Hybrid Genetic Algorithm and Cellular Automata Combined Traffic Signal Control and Route Guidance

Signal control and route guidance jointly influence traffic flow in time and space. Over the last three decades, combined traffic signal control and route guidance (CTSCRG) has been research emphasis. Firstly, the conceptual structure and definition of CTSCRG was analyzed. Then, the mathematical models of CTSCRG were summarized. Link travel time function and signal control policy have significant influence on solution uniqueness and convergence of CTSCRG model. Simulation-based method can allow more complex interactions, therefore win in reality value than travel time formula. The paper combines hybrid genetic algorithm with cellular automata simulation to calculate travel time and optimize signal setting plan. Iterative simulation and assignment procedure is built: road is discretized by cellular automata. Traffic flow dynamics is represented by cell transmission model; signal setting is optimized by hybrid genetic algorithm; vehicle agent can receive route guidance information and select suggested route. The simulation result is encouraging, when combined traffic signal control and route guidance equilibrium converge, the saving in total travel time is 54.4%.

Revised MAXBAND Model for Bandwidth Optimization of Traffic Flow Dispersion

MAXBAND model of bandwidth optimization assumes that all the vehicles have the same speed. But there is traffic flow dispersion because of different vehicle performance. Traffic flow dispersion has been described by normal distribution and geometric distribution. The paper integrates traffic flow dispersion model into MAXBAND model to optimize bandwidth of traffic flow dispersion. Revised model will more effectively optimize bandwidth.

Vehicular Traffic Flow Dispersion Prediction of Coordinated Signal Control Based on Support Vector Regression

Signal control can reduce traffic delay and improve traffic safety. But when the vehicle meets with red light, it must stop at intersection to wait green light. Coordinated signal control can improve the continuity of vehicular traffic flow movement and reduce delay. Coordinated signal setting is based on platoon dispersion prediction. Thus, improving prediction accuracy can obtain significant benefit for signal coordination. The paper bases on support vector regression to predict platoon dispersion and compares prediction accuracy with Robertson formula.

Traffic Analysis and OD Travel Time Matrix Based on Two-Fluid Model

The traditional four-step traffic demand forecast model is limited by the long processing time and high cost of origin destination (OD) matrix investigation. A dynamic traffic assignment model is difficult to establish for a large-scale network. To obtain the traffic character and quickly search the shortest path of different zones for dynamic route guidance, the authors analyze travel distance and travel time using taxi global positioning system (GPS) data, fit the travel time and stop time relation curve, and propose a method of traffic analysis using the two-fluid curve. The authors find that the bandwidth of the two-fluid curves is valuable for traffic operation and guidance. Subsequently, the authors analyze the relationship between unit distance travel time and unit distance stop time of different zones. The results verify that the sensitivity of different traffic zones varies. Finally, matrix iteration is used to calculate the shortest travel time path under different unit distance stop times, and the OD travel time matrix is analyzed. The findings indicate that the two-fluid method can be used for dynamic route guidance.

Deriving the Macroscopic Fundamental Diagram for an urban area using counted flows and taxi GPS

This paper applies the Macroscopic Fundamental Diagram (MFD) and the Generalized Macroscopic Fundamental (GMFD) Diagram to estimate traffic characters for an urban area. We used traffic flow data manually counted from intersection traffic video and taxi GPS data to derive MFD and GMFD. The method is suitable to recognize traffic situation. These figures show the property of the road network, which consists of infrastructure and control. Mastering the properties of the network supply side, we can better manage and control traffic demand. In the process of data analysis, we found that the time step for data processing is a determinant for the MFD and GMFD shapes. That means different time step produces different results.