LUTI (Land Use and Transportation Interaction) models are decision-making aid tools to simulate complex dynamic bilateral feedback between transportation and land-use models within a territory. LUTI models appraise several further planning scenarios to arrive at the most appropriate decisions.
Making decisions based on the models that are not calibrated or calibrated properly might be misleading and even incorrect. Although calibration (parameter estimation) is a crucial requirement of LUTI models, fully automated approaches using multi-objective functions have not been fully addressed. There is no standard procedure for LUTI model calibration. Modelers instead use conventional techniques to calibrate a specific element of a model or estimate a group of model parameters with little or no concern for a global scheme.
This thesis aims to develop a fully automated global calibration approach using multi-objective functions. In order to overcome this constraint, a novel calibration methodology is introduced for the parameters of the land-use model, using a Differential Evolution (DE) algorithm. A global sensitivity analysis was performed to identify the most critical land-use model parameters. These parameters were then calibrated using the differential evolution algorithm with the Root Mean Square Error (RMSE) and Mean Absolute Normalized Error (MANE) as standard statistical metrics to measure the goodness of the proposed calibration approach. The proposed technique (DE algorithm) offers five critical capabilities for calibrating LUTI models: 1) global estimation, prioritizing over local estimation, 2) accommodating multi-objective functions, 3) continuously enhancing results, 4) easy adaptability, and 5) incorporation of multiple parameters in the calibration process. The performance of the proposed calibration technique was assessed using the TRANUS land-use model. The approach was validated and consolidated, evaluating convergence, error minimization, and the ratio between modeled and observed data. These assessments involved comparisons with two established optimization techniques: Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). Our experiments indicate that employing the Differential Evaluation algorithm resulted in the proposed approach outperforming both GA and PSO techniques. The Differential Evaluation algorithm provided superior performance and demonstrated excellent stability and diversity in solutions.
|
