Study On Methods And Applications For Process-Oriented Dynamic Decision Making

Today’s decisions are becoming more complex, with greater uncertainty, more rapidly changing conditions and increasing time pressure. Since the essence of human survival environment is dynamic, relative to the static decision making, dynamic decision making more universal, but the decision theory currently widely used is largely developed in the static case, the lack of research on the decision making process, can not accurately describe the behavior of individual decision making in dynamic, real-world environment. Therefore, the study on methods and applications for the process-oriented dynamic decision making should become an important content of decision theory, and it will present huge significance of theory research and value of engineering application.In terms of the process-oriented dynamic decision making method studied in the paper, modeling and simulation analysis of the decision making process of individual cognition, judgment and decision making behavior are carried out, to explore the psychological evolution process of individual decision making behavior changes over time in a dynamic, uncertain and time pressure environmental. Contributions of the dissertation are as follows:Study of the multi-attribute dynamic decision making modeling approach based on the decision field theory. On the basis of analysis of multi-attribute dynamic decision making process, modeling method is elaborated. The model is used to explain and simulation the cognitive processes of multi-attribute dynamic decision in a dynamic, uncertain and time pressure environmental, and predict the choice probability and choose preferences change with time. A dynamic decision making framework is presented.Research on preference reversals in multi-attribute decision making. The paper focuses on the two forms of preference reversal phenomenon in multi-attribute decision making:(1) different preferences measurements (choice and match) lead to preference reversal;(2) to add (or take off) one option in a set of options cause preference reversal. For the first type of preference reversal phenomenon, the mathematical analysis is made based on multi-attribute dynamic decision matching model and computer simulation algorithm. For the second type of preference reversal phenomenon, the similarity effect, the attraction effect and the compromise effect are quantitatively explained using multi-attribute dynamic decision making model based on decision field theory.Developing a dynamic route choice model for drivers’ route deliberation processes. The framework with process-oriented, uncertainty, stochastic, descriptive and dynamic properties is proposed, and the interactions of driver psychology, road conditions, and decision-making time is taken into account and the travel time, distance and the number of intersections is set as the main attributes in the model, so that the dynamic route choice model is developed to capture the psychological process involved in route choice decisions for varying degrees of uncertainty as well as time pressure.Developing a route choice model for drivers’ pre-trip and en-route choice behavior under guidance information. Based on Bayesian theory, a road conditions dynamic updating model is presented in light of the guidance information and the driver’s previous travel experiences. Then, the route choice behavior model under guidance information is formed by the fusion of the process-oriented vehicle dynamic route choice model and the road conditions dynamic updating model. An en-route driver route choice behavior model that uses concepts from Decision Field Theory and Bayesian belief network is proposed. A real-time planning algorithm for route choice processes is discussed in great detail. Critical factors that affect drivers’ response to real time traffic information are quantitatively studied through interactive simulation.Computer simulated route choice experiments are designed to monitor and record route choice behavior data under varying experimental conditions for analyzing drivers’ route choices and for the development and calibration of an operational route choice model. Estimation of the route choice model parameters is performed based on the experimental observations. Genetic algorithms are used as the optimization tool to calibrate model parameters and minimize the discrepancy between model output and observed behaviour. Finally, an operational route choice model for drivers’ pre-trip and en-route choice behavior is developed.