Accurate control of driving system for crawler plant protection robot (CPPR) is a challenging problem in smart agriculture. The soil is usually soft and mixed with straw and weeds in field. When the CPPR is in contact with the soil, uncertain shear disturbance will occur due to pressure subsidence and shear action. In serious cases, it may even lead to failure, which significantly increases the control difficulty. In this paper, a fractional order sliding mode fault tolerant (FSMFT) control based on online real-time fault estimation is proposed. Firstly, the fault model of the driving system is established, and a fault observer based on nonlinear fault observer (NFO) is designed according to the fault model, which can estimate the fault information of the driving system in real time online. Then, a fractional order sliding mode (FOSM) fault tolerant control method is proposed according to the observation information of the observer. It will be shown that by incorporating fault tolerant control theory and the observer technique, the proposed control can ensure a good control performance in terms of tracking precision, robustness, and fast convergence, etc. Finally, simulation experiments are demonstrated to show the superiority of the proposed control algorithm.