Local path planning for autonomous vehicles based model predictive control using velocity obstacles potential field in emergency situation

Abstract

Recently, as the demand and interest for autonomous vehicles increase, many studies are being conducted. The operating mechanism of autonomous vehicles can be largely classified into perception, decision-making, and control. In the three areas, particularly in the area of decision-making, no clear problem definition or solution was provided. Decision-making can be divided into global path planning, behavior selection, and local path planning. In this paper, we focus on local route planning in emergency situations such as collision with dynamic obstacles. So, many approaches have been made to use the Model Predictive Control(MPC), which has a vehicle model as a constraint. But, Few studies have used models that reflect kinetic characteristics in the longitudinal direction. In this paper, we establish a dynamics model that has propelling torque, brake torque and steering angle of electric autonomous vehicles as control inputs. Using this, we solve the potential field method using velocity obstacles(VO) for efficient collision avoidance in emergency situation such as low Time to Collision(TTC). It is shown that the velocity obstacles based potential field method can effectively handle the situation where collision with dynamic obstacle is expected while satisfying the vehicle dynamic characteristics.