This thesis develops a robust output-tracking control strategy for autonomous rover systems with redundant control directions. Taking advantage of their redundancy, an optimal distribution of control actions is proposed to enhance the dynamic traction of such systems. A robust optimal output-tracking control strategy for underactuated mechanical systems subject to mixed holonomic and nonholonomic constraints is presented. Then, we develop a novel methodology to optimally lift the proposed robust control law from the output dynamics to the space of control actions. This methodology aims at enhancing the dynamic traction of autonomous rovers, without affecting the tracking performance of the system. To improve approximations used in this optimization, a disturbance observer is designed in the time domain. The developed control strategy is evaluated for a six-wheel autonomous Lunar rover in a simulation environment. The developed traction control algorithm is also implemented in a software-in-the-loop simulation environment using Vortex Studio.