Trajectory planning aims at computing an optimal trajectory through theminimization of a cost function. This paper considers four dierent scenarios: (i)the rst concerns a given trajectory on which a cost function is minimized by aacting on the velocity along it; (ii) the second considers trajectories expressedparametrically, from which an optimal path and the velocity along it arecomputed; (iii), the case in which only the departure and arrival points of thetrajectory are known, and the optimal path (in the sense of minimizing a givencost function) must be determined; and nally, (iv) the case involving uncertaintyin the environment in which the trajectory operates. When the considered costfunctions are expressed analytically, the application of Euler-Lagrange equationsconstitutes an appealing option. However, in many applications, complex costfunctions are learned by using black-box machine learning techniques, for instancedeep neural networks. In such cases, a neural approach of the trajectory planningbecomes an appealing alternative. Dierent numerical experiments will serve toillustrate the potential of the proposed methodologies on some selected use cases.