Nowadays it is eagerly expected that, on one hand the environmental impact of agricultural activities is decreasing, and on the other hand the level of production is increasing in order to match the consumption demand of the growing worldwide population. To meet these opposite expectations, new production tools have to be developed. Recent advances in off-road mobile robotics may bring to promising solution in order to address such a problematic. Since large area coverage, as well as high flexibility, are more particularly looked for in this paper, multi-robot cooperation for field operations is investigated. The proposed framework relies on a path tracking approach dedicated to the formation control of several robots. In an off-road context, because of bad grip conditions, classical control algorithms cannot be used straightforwardly. Moreover, low level delays and stability issues (rollover risk or obstacle collision) have also to be taken into account. Adaptive and predictive control laws are here designed to achieve accurate motion control of each robot, despite highly varying and unpredictable contact conditions. The multi-robot configuration is imposed by defining set points in terms of lateral deviations and longitudinal distances between each robot. In addition, robot safety is addressed through a traversability evaluation within the vicinity of each robot. Finally, several alternative paths are constantly considered and the optimal one is decided with respect to robot stability and maximal velocity criteria, assessed from a numerical terrain model.