Ionic polymer-metal composites (IPMCs) are promising candidates in various sensing and actuation applications due to their light weight, large bending, and low actuation voltage requirements. However, IPMCs are still in the early stage of development, and their bending response can vary widely depending on various factors such as fabrication process, water content, temperature, and contact with electrodes. To control IPMCs in a predictable manner and to minimize the effects of plant uncertainty and external disturbances, a precise and robust control scheme is required. In the present work, a three-part adaptive feedforward control architecture is employed for IPMC deflection control. First, adaptive identification is performed to identify changes in the dynamic behavior over time and in the input voltage using a gradient descent method. Second, an adaptive feedforward controller is implemented to control the dynamic response of the plant, where the IPMC displacement is observed and is used to adjust the parameters of the controller. Third, noise and disturbance cancelling is performed using an additional adaptive canceller, which does not affect the system dynamics. Our results show that the adaptive identification and feedforward controller with disturbance cancellation using the gradient descent method provides accurate tracking performance under plant variation and disturbance. Especially, the fast convergence speed of the proposed technique makes it practical for online control.