In this work, an energy planning strategy is proposed for over-actuated unmanned road vehicles (URVs) having redundant steering configurations. In fact, indicators on the road geometry, the actuation redundancy, the optimal velocity profile, and the driving mode are evaluated for each segment of the URV's trajectory. To reach this objective, a power consumption estimation model is developed for the URV. Due to the presence of unknown dynamic parameters of the URV and uncertainties about its interaction with the environment, an artificial intelligence (AI) technique, based on data-learning qualitative method, is used for the power consumption estimation, namely Adaptive Neuro Fuzzy Inference System (ANFIS). The ANFIS model is obtained using trained data from a Real URV dynamics. Then, an energy digraph is built with all feasible configurations taking into account the kinematic and dynamic constraints based on a 3D grid map setup, according to velocity, arc-length, and driving mode. In this weighted directed graph, the edges describe the consumed energy by the URV along a segment of a trajectory. The vertices describe the start and end points of each segment. Subsequently, an optimization algorithm is applied on the digraph to get a global optimal solution combining driving mode, power consumption, and velocity profile of the URV. The obtained results are compared with the dynamic programming method for global offline optimization. Finally, the obtained simulation and experimental results, applied on RobuCar URV, highlight the effectiveness of the proposed energy planning.