주변 차량과 도로 모델을 고려한 모델 예측 방식의 횡 방향 제어

Abstract

This paper proposes potential field based lateral planning and control method using model predictive approach for autonomous driving. The potential field is constructed as a superposition of disparate functions of surrounding vehicles, road boundaries and lanes. Vehicle states related to comfort and safety are formulated with model predictive approach and a cost function is designed. Then based on the potential field model, the potential risk using predicted states is added to the cost function above. With the combined cost function and roughly discretized input set, the suboptimal solution which has the smallest value of the cost function for path planning and control is solved. The proposed algorithm is developed with MATLAB/Simulink and its performance is evaluated using the commercial software CarSim.