Human–robot interaction is inherently available and used actively in ankle rehabilitation robots. This interaction causes disturbances to be counteracted on the rehabilitation robots in order to reduce the side effects. This paper presents a fractional order proportional–integral–derivative controller to improve the trajectory tracking ability of a developed 2-degree of freedom parallel ankle rehabilitation robot subject to external disturbances. The parameters of the controller are optimally tuned by using both the cuckoo search algorithm and the particle swarm optimization algorithm. A traditional proportional–integral–derivative controller, which is also tuned using both of the algorithms, is designed to test the performance of the fractional order proportional–integral–derivative controller. The experimental results show that the optimally tuned FOPID controller improves the tracking performance of the ankle rehabilitation robot subject to external disturbances significantly and decreases the steady-state tracking errors compared to the optimally tuned PID controller.