Title:An Integrative Data-Driven Physics-Inspired Approach to Traffic Congestion Control

Abstract:This paper offers an integrative data-driven physics-inspired approach to model and control traffic congestion in a resilient and efficient manner. While existing physics-based approaches commonly assign density and flow traffic states by using the Fundamental Diagram, this paper specifies the flow-density relation using past traffic information recorded in a time sliding window with a constant horizon length. With this approach, traffic coordination trends can be consistently learned and incorporated into traffic planning. This paper also models traffic coordination as a probabilistic process and obtains traffic feasibility conditions using linear temporal logic. Model productive control (MPC) is applied to control traffic congestion through the boundary of the traffic network. Therefore, the optimal boundary inflow is assigned as the solution of a constrained quadratic programming problem.