Hyper-redundant robot manipulators possess a very large degree of kinematic redundancy and are capable of motion similar to that of snakes and elephant trunks. Because of the computational burden required to calculate the pseudo-inverse of the manipulator Jacobian matrix for high degree of freedom robots, hyper-redundant manipulators have proven challenging to control by traditional methods. Additionally, control can be further complicated because the large number of joints and links in a hyper-redundant arm can be a source of measurement error in real-world systems. A fuzzy logic and neural network based control system for hyper-redundant arms is presented which operates on data from real-time vision. The neural network maps goal position and orientation to desired arm configuration. A modified region fill algorithm is used to provide an estimate of the current configuration as seen in two camera views. A fuzzy logic rule base constructed from human intuition specifies motor signals which servo the arm from the current position to the goal position according to velocity profiles modeled after human goal-directed movement strategies. As a test case, the control system is applied to a thirty-two degree of freedom robot arm designed and built at Rice University. The controller is demonstrated to provide accuracy similar to that of humans on certain tasks. Although application is made specifically to a hyper-redundant arm, the control system developed in this research also could be applied to many lower degree of freedom manipulators, provided that their motion is heuristically easy to describe.