Real-Time Unmanned Aerial Vehicle-Based Traffic State Estimation for Multi-Regional Traffic Networks

Abstract

Traffic state estimation is a challenging task because of the collection of sparse and noisy measurements from fixed points in the traffic network. Induction loops, as they are non-intrusive, can observe any area of the traffic network on demand and provide accurate traffic density and speed measurements. Our main contribution is the development of an optimization framework where small parts of the traffic network are monitored by Unmanned Aerial Vehicles (UAVs) and accurate estimates of traffic density and mean speeds for every region in the traffic network are returned in real-time. Assuming regional-based traffic dynamics, a cyclical UAV flight path is defined for each region. One UAV is assigned to each flight path and monitors a small area of the region below. The UAV-based traffic measurements are expressed as moving averages to smooth out fluctuations in traffic density and mean speed. A moving horizon optimization problem is formulated, which minimizes the estimation and process errors over a moving time window. The problem is non-convex and challenging to solve, because of the presence of nonlinear traffic dynamics. By considering free-flow conditions, the optimization problem is recast to a quadratic program that returns density estimations for each region of the traffic network in real-time. Simulation results compare our UAV framework to an alternative, where the whole traffic network is monitored by UAVs. Both frameworks obtain similar results, despite the alternative framework using more UAVs than our framework.