The Mars Exploration Rover Spirit will henceforth be a stationary science platform, as its current mobility capabilities have been unable to free it from a sand trap for several months. Mobility systems of future planetary exploration robots will need to reduce risk of entrapment, access more challenging terrain, and cover more ground than ever before. This work presents a novel center-of-gravity (CoG) control algorithm, designed to increase rover mobility over step obstacles. The control algorithm is applied to a mobile robot platform featuring a reconfigurable chassis / suspension system. The 6-wheeled platform is equipped with active wheel-walking degrees of freedom in addition to driving, steering, and passive suspension DoFs typical for such robots. The controller is based on a new concept introduced here, the Contact-Angle Adjusted Support Plane, and uses wheel-ground contact angles to place the CoG such that weight acting on wheels encountering difficult local terrain is reduced. Mobility analysis, using the detailed multibody dynamics simulator RCAST (Rover Chassis Analysis and Simulation Tool), shows that the control algorithm dramatically improves step-climbing ability. The CoG controller can increase the height of obstacle surmountable by a factor of 2 to 3. Wheel-step interaction modes, joint angle limits for wheel-walking DoFs, and other practical mechanical considerations are discussed in the context of a hardware prototype of the studied chassis design.