Automotive longitudinal speed pattern generation with acceleration constraints aiming at mild merging using model predictive control method

Abstract

To ensure safety and simplicity in merging path generation for a realistic reliable and mild merging, this paper proposes a merging path generation method. In the proposed method, the merging problem is considered in two-dimensional space and formulated into a one-dimensional space optimization problem by relating the longitudinal motion of the merging vehicle to the lateral motion of it. In this way the optimization problem would be much simpler and therefore the computational time could be shorter than formulating it into a two-dimensional problem. Moreover, the parameters are chosen appropriately so that the variation of the acceleration of the main lane vehicle is less severe than that of the merging vehicle, which is consistent with the practice. To realize mild merging, the merging path is optimized while the accelerations of the relevant vehicles are optimized through the model predictive control (MPC) method. With the proposed method, the merging vehicle can merge smoothly and realistically in cooperative with the main lane vehicle. The effectiveness of this method is verified by a computer simulation of the motions of one merging vehicle and one main lane vehicle. The initial conditions of the merging are set realistically according to the data drawn from actual merging scenes. The results proved that, with the proposed method the merging vehicle can merge mildly in cooperation with the main lane vehicle.