A vehicle with four powered caster wheels can provide much more motion flexibility in a constrained environment. However, the dynamic modelling and control of such system is challenging due to its high redundancy. This paper investigates the dynamic model and tracking control for a four-powered-caster vehicle (4-PCV) on complex terrain without any additional sensor. The torques applied to the wheels are dynamically redistributed based on the real-time conditions of the whole wheel–ground interactions so that the vehicle can track the desired trajectory when moving on different terrains. A dynamic model considering the wheel–ground interaction is first derived. Then a novel approach based on a probability scheme is proposed to identify the status of the vehicle and the wheel slip ratio by only observing the velocity feedback from motors encoders. Based on this real-time perception information, a tracking controller and a torque distribution scheme are applied to make sure that each wheel can be self-adapted to meet a complex wheel–ground condition to eliminate or reduce the probability of slippage. The effectiveness of the proposed estimation approach and the performance of the torque distribution schemes are verified by simulation.