| Summary |
Po - Kai Peng，2007.06
Graduate Institute of Transportation Science, Tamkang University
Dynamic programming (DP) determines the optimum solution to an n-variable problem by decomposing it into n stages with each stage constituting a single variable problem. This technique can be easily constructed to solve a shortest-path problem. If stochastic nature of problem is concerned, probabilistic dynamic programming can be applied in that the states and the returns at each stage are probabilistic. In this thesis, a route guidance problem is solved as finding a dynamic shortest route in a time dependent probabilistic programming problem. Under this formulation, route choice probability is introduced at each consecutive decision node (stage of diversion). The series of choices over entire trip can be defined as node-to-node dynamic route choice behavior.
The node-to-node dynamic route choice behavior is of the most interest to study the individual driver’s route choices under the influence of the route guidance information where individual driver makes consecutive route switch decisions along with the traveling route. This particular issue has been successfully modeled with various forms and extensions under the notion of the “Indifference Bands” applied with Probit model specifications by Tong and his students at Tamkang University in recent years. The probability of “swithching” or “route choice” at each decision node along the route, reflecting the compliance outcome to the routing diversion via either in-vehicle devices or road side VMS, can therefore be estimated under these model specifications.
This thesis applies a newly developed network simulation program, DynaTAIWAN, for generation time dependent system performance indices over the selected study area (e.g., link travel time) with the embedded dynamic route assignment procedure. A complimented survey with rolling-plane feature was designed to perform controlled experiments where selected sample of travelers were selected to perform routing decision over trips for simulated scenarios under either In-Vehicle guidance or VMS environment.
Choice models were calibrated and probabilistic dynamic programming formulated accordingly. Expected optimum route was then solved for each individual sample traveler respectively. The results have demonstrated the possibility of various routing suggestions across individual driver departing at the same space-time slot, which suggested the development of the diversified dynamic route guidance based on the current modeling treatments and findings. The analysis of aggregate behavior over entire network can be encouraged in the future. |
