Soft robots, robots that are constructed out of soft materials or using compliant actuation methods, can operate safely in complex environments without fear of damaging their surroundings or themselves. However, the soft materials and structures can be imprecise and difficult to control. This thesis seeks to remedy this problem for reverse pneumatic artificial muscle (rPAM) actuators, which can apply force by extending when pressurized. First, we discuss the modeling of simple linear rPAMs as well as the motion control of a linear rPAM-driven 1-degree-of-freedom (DoF) revolute joint. Control is done through an iterative sliding mode controller with and without a feedforward term. Next, we adapt this control approach to a soft planar bending segment and use a model reference adaptive controller to compensate for the variability of soft material mechanical properties. From there, we expanded the iterative sliding mode controller and used it to control 2-DoF joint modules using three linear rPAM actuators. We subsequently replaced this universal joint with the human wrist and developed a system to provide haptic feedback to the users, improving their performance computerized line-following task. We then developed a model to predict the behavior of bending actuators, both under load and while pressurized internally, which we used to perform inverse kinematic path following. These techniques represent a meaningful advancement in understanding and improving soft actuators, allowing them to move with speed and precision while resisting external forces.