Optimization-Based Path Planning Algorithm with Finite Traversable Area

Path planning is one of the important technologies constituting an autonomous navigation system, aims to create a path from a starting point to a destination point. The generated path cannot reflect various environmental changes that may be encountered during actual driving because all information for the path generation cannot be obtained in advance. Therefore, although an autonomous ground vehicle (AGV) eventually follows the resulting path, a new path considering the situation around the AGV is required. In this paper, we propose a path planning algorithm that handles local environmental information around the vehicle by utilizing a finite horizon optimal control. In order for that, the proposed method is formulated as an optimization problem which is composed of a cost function, vehicle dynamic characteristics, control input constraints, and environmental variation around the AGV. In particular, this study views the changing environment during the navigation as traversable or non-traversable area and defines the area as an analytic potential function. Then, the function is incorporated into the cost function of the path planning problem and its optimal solution is obtained using the first-order gradient descent method. To validate the performance of the proposed method, numerical simulations are conducted with a virtual environment autonomous navigation system composed of CarMaker and robot operating system (ROS), and the results are analyzed.