Mobile robotic constitutes a promising way to reduce the environmental impact of agricultural activities and allow to keep the level of production to satisfy the growing population demand. In previous work, a formation control law, accurate despite typical off-road conditions (low grip, terrain irregularities, etc), based on a nonlinear observer-based adaptive control was presented. Satisfactory advanced results have been reported in lateral servoing but inaccuracies in longitudinal regulation have been notived. In this paper a new predictive approach dedicated to both longitudinal and lateral servoing is porposed. We consider mobile robots made of a single body accounting for bad grip condition thanks to a observer based approach. After presenting the modelling of the formation and the control law used to maintain longitudinaland lateral deviation with respect to the reference trajectory, we introduce the predictive approach on the velocity. Full-scale experiments demonstrate the performance of the proposed approach.