Study on the formulation of vehicle merging problems for model predictive control

Abstract

Vehicle-to-vehicle or road-to-vehicle communication has been used to control automated vehicles in many studies. However, the more expensive the required road and vehicle facilities are, the slower the spread of automated vehicles will be. Therefore, this paper proposes a method to formulate a merging problem for model predictive control (MPC). To this end, information obtained from inexpensive in-vehicle cameras is used to realize more affordable automated vehicles. We proposed the use of sigmoid curves to model merging roads. The advantage of using a sigmoid curve is the stability of the calculation and the ability to model the merging road with minimal information. This study models the road by detecting merging points from onboard camera images. A neural network was used to estimate the speed of the mainline vehicle. By estimating the speed, it is possible to estimate the position of the mainline vehicle one step in the future. This means that the merging vehicle can merge without colliding with the mainline vehicle. The proposed method was used to simulate the quality point model and shown to solve the merging problem on a parallel merging road, where the mainline vehicles travel straight ahead.