Ionic polymer metal composite (IPMC) is used in various bio-inspired systems, such as fish and tadpole-like robots swimming in water. The deflection of this smart material results from several internal and external factors, such as water distribution and surface conductivity. IPMC strips with a variety of water concentration on the surfaces and surface conductivity show various deflection patterns. Even without any external excitation, the strips can bend due to non-uniform water distribution. On the other hand, in order to understand the effects of surface conductivity in an aquatic environment, an IPMC strip with two wires connected to two distinct spots was used to demonstrate the power loss due to the surface resistance. Three types of input signals, sawtooth, sinusoidal, and square waves, were used to compare the difference between the input and output signals measured at the two spots. Thick (1-mm) IPMC strips were fabricated and employed in this research to sustain and drive the robot with sufficient forces. Furthermore, in order to predict and control the deflection, researchers developed the appropriate mathematical models. The special working principle, related to internal mobile cations with water molecules, however, makes the system complicated to be modeled and simulated. An IPMC strip can be modeled as a cantilever beam with loading distribution on the surface. Nevertheless, the loading distribution is non-uniform due to the non-perfect surface metallic plating, and four different kinds of imaginary loading distribution are employed in this model. On the other hand, a reverse-predicted method is used to find out the transfer function of the IPMC system according to the measured deflection and the corresponding input voltage. Several system-identification structures, such as autoregressive moving average with exogenous (ARX/ARMAX), output-error (OE), Box-Jenkins (BJ), and prediction-error minimization (PEM) models, are used to model the system with their specific mathematic principles. Finally, a novel linear time-variant (LTV) concept and method is introduced and applied to simulate an IPMC system. This kind of model is different from the previous linear time-invariant (LTI) models because the IPMC internal environment may be unsteady, such as free cations with water molecules. This phenomenon causes the variation of each internal part. In addition, the relationship between the thickness of IPMC strips and the deflection can be obtained by this concept. Finally, based on the experimental results above, an aquatic walking robot (102 mm × 80 mm × 43 mm, 39 g) with six 2-degree-of-freedom (2-DOF) legs has been designed and implemented. It walked in water at the speed of 0.5 mm/s. The average power consumption is 8 W per leg. Each leg has a thigh and a shank to generate 2-DOF motions. Each set of three legs walked together as a tripod to maintain the stability in operation.