**2.6 Path planning**

In this task, the main goal is to find a geometric path from an initial point to an endpoint. Sometimes, the vehicle dynamics can be considered in the problem even though this will mean the work can only be applied to vehicles with the same characteristics [22, 23]. A better approach is to work in path planning considering a general solution that is not tied to any specific vehicle [24]. There are two main approaches global route planning and local path planning. Global planners search routes from origin to the final destination, some proposals focus on efficiency in real-time traffic. In contrast, others can compute directions in milliseconds, and others consider space requirements. Local planners find trajectories in real-time considering obstacles, and their objective is to complete the global route. Despite the different approaches, there is an existing controversy among some researchers that if an AV should drive like a human or should look for the optimum path.

Some proposals use SVM [22] and Genetic Algorithms (GA) [25]; with these algorithms usually, other methods are applied in the first step, for instance, A\* algorithm which is a typical graph search algorithms in pathfinding. Another method extensively used is the use of Artificial Neural Networks (ANN). There are different variants, the more used in autonomous driving include CNN [23] and Fully Convolutional Networks (FCN) [24].
