Dynamic Estimation of Headway Distance in Vehicle Platoon System under Unexpected Car-following Situations

Abstract

The headway distance between vehicles in a platoon is difficult to measure unless the vehicles are equipped with costly equipment such as laser or radar sensors, or an image processing system. This paper proposes an alternate approach to estimating headway distance indirectly from measurable variables such as the acceleration rate and velocity of selected vehicles in the platoon. Assuming a six-vehicle platoon, a particle filter (PF) and an unscented Kalman filter (UKF) are applied in order to estimate the headway of each vehicle based on the measurement variables of three (or all) vehicles in the platoon. The state-space models of the PF and the UKF are given by the conservation equation of headway and the conventional car-following model. To evaluate the PF and UKF performance, two scenarios were prepared: one assumed that prior knowledge of a model parameter differed from what was actually observed as posterior information, the other was a situation where a platoon vehicle slowed down unexpectedly during the car-following process. In both situations, the state-space model itself was unable to describe the dynamics of headway distance precisely, and the PF and the UKF were applied to minimize the headway distance errors caused by the incorrect model parameter and the unexpected vehicle slowdown. Numerical analysis demonstrated that both the PF and UKF were successful in estimating the headway distance, even when the car-following model did not express the true car-following phenomena. We also determined that, if all vehicles are equipped as probe cars, and thus capable of measuring the acceleration rate and velocity, the PF is superior to UKF when estimating the headway distance precisely. However, UKF is more stable than the PF when measurements are not taken from all vehicles, especially when a vehicle in the platoon unexpectedly slows downs during the car-following process.