Loading...
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 51019 (09)
- University: Sharif University of Technology
- Department: Civil Engineering
- Advisor(s): Samimi, Amir
- Abstract:
- Real-time prediction of traffic flow and vehicle composition in urban network is the main purpose of this research. A two-step model is presented as an alternative for costly methods of data collection. At the first step, Passenger Car Equivalent of flow is estimated by an Artificial Neural Network model, and then a Multiple Discrete-Continuous Extreme Value model defines vehicle composition. Real-time speed data from Google-Maps API, land-use information, and transportation network specifications of Tehran are used. The observed traffic from 95 Automatic Number Plate Recognition cameras are gathered from the CBD zone of Tehran on spring 2017. Goodness of fit and Root Mean Square Error of the calibrated models are reported. PCE prediction model has 63% goodness of fit. Vehicle composition model can respectively predict passenger car, taxi, pickup and heavy vehicle types by 91.7%, 69.4%, 40.9%, 32.5% fit. Weighted average of the fit for vehicle composition is 81.9%. Goodness of fit in consequent modeling of traffic flow based on vehicle composition are 58.3, 57.5, 31.7, 28.5 and 56.4 respectively for passenger car, taxi, pickup, heavy vehicles and weighted average
- Keywords:
- Artificial Neural Network ; Real-Time Estimating ; Multiple Discrete-Continuous Extreme Value Model (MDCEV) ; Passenger Car Equivalent ; Vehicle Composition
- محتواي کتاب
- view