Autonomous robot design is a challenging field in artificial intelligence. It is not easy to design autonomous robots to perform useful tasks in real environments. There are two reasons why designing a control system for an autonomous robot is a difficult task. The first reason is that it is difficult to coordinate different parts of the robot to work together and to operate as the control system requested. The second reason is that autonomous robots interact with an environment, which means they may face unpredictable environment situations so that their control system may be misled. A novel approach using Genetic Algorithm simulation is discussed in this thesis and the results of the simulation are verified physically by real robot implementation.

Autonomous robot design is a challenging field in artificial intelligence. It is not easy to design autonomous robots to perform useful tasks in real environments. There are two reasons why designing a control system for an autonomous robot is a difficult task. The first reason is that it is difficult to coordinate different parts of the robot to work together and to operate as the control system requested. The second reason is that autonomous robots interact with an environment, which means they may face unpredictable environment situations so that their control system may be misled. The control system must send out its commands in real-time, and each command affects not only how well the current goal is completed but also the situation of the robot in the environment. For these reasons, a robot that works well in simulation may not behave well in a real environment.