This paper analyses a mobile robot with independently rotating wheels travelling on uneven but smooth ground, including ascending or descending surfaces. We formulate a mathematical expression for the energy cost of the robot's movement. For our analysis, we utilise the principle of virtual work and assume that the robot moves with a fixed arrangement of wheel axes and without using a steering handle. The mathematical model reveals that the coefficient of friction and the payload distribution dominate the wheel behaviour, including slipping and skidding. We minimise the virtual work expression to determine the robot's motion complying with driven wheels. The model also enables us to estimate trajectories for different ground conditions. A hybrid robot, PEOPLER-II, is used to demonstrate the predicted motions, including turns and spins, by following angular velocity control rules. Experimental data verifies that the proposed formulation and minimisation of virtual work are valid techniques for predicting a robot's trajectory. The method described is widely applicable to wheeled robots having independently driven wheels.