MechaCells are designed as closed, scalable and modular semi-autonomous devices that can be used alone or part of a pack. In this paper, we discuss a locomotion system that uses the reaction force produced by a rotating unbalance that moves in a spherical domain with a steering mechanism. In order to produce the precise motion capability, a multi-loop controller is developed. This controller uses a friction compensation algorithm based on the mathematical model of the locomotion system. To improve the accuracy of tracking, conventional LuGre friction estimation model is extended for rapid directional changes of the MechaCell during planar motion. The linear and rotational acceleration of the device is also included in controller calculations since it affects the locomotion force generated by the unbalanced mass. The resulting control system is validated both with simulations and experiments and the effectiveness of the extended model and the controller is verified. Our results show significant improvement when a detailed friction compensation observer is used in the controller that includes the effect of sudden steering changes for precise path following.