This chapter presents two different approaches of designing genetic-based controllers for an unstable physical system (a simulated pole-cart system). One approach induces rule-base controller using a simple genetic algorithm (GA) and the other evolves neuro-controller applying a recently developed Structured Genetic Algorithm (sGA) which appears to offer improvements over a simple GA approach. The control task here is a typical unstable, multi-output, dynamic system in which a pole is supposed on a controllable cart, and the controller must keep the pole upright (within a specified vertical angle) and the cart within the limits of the given track. In this chapter, we first describe a simple GA based learning method for inducing control rules, and then demonstrate the evolvability of neuro-controller using a Structured GA.