eduzhai > Applied Sciences > Engineering >

Highway Traffic State Estimation Using Physics Regularized Gaussian Process Discretized Formulation

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 26 pages

Abstract: Despite the success of classical traffic flow (e.g., second-ordermacroscopic) models and data-driven (e.g., Machine Learning - ML) approaches intraffic state estimation, those approaches either require great efforts forparameter calibrations or lack theoretical interpretation. To fill thisresearch gap, this study presents a new modeling framework, named physicsregularized Gaussian process (PRGP). This novel approach can encode physicsmodels, i.e., classical traffic flow models, into the Gaussian processarchitecture and so as to regularize the ML training process. Particularly,this study aims to discuss how to develop a PRGP model when the originalphysics model is with discrete formulations. Then based on the posteriorregularization inference framework, an efficient stochastic optimizationalgorithm is developed to maximize the evidence lowerbound of the systemlikelihood. To prove the effectiveness of the proposed model, this paperconducts empirical studies on a real-world dataset that is collected from astretch of I-15 freeway, Utah. Results show the new PRGP model can outperformthe previous compatible methods, such as calibrated physics models and puremachine learning methods, in estimation precision and input robustness.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×