| Abstract: | Traditionally, the task of developing control architecture for a situated autonomous robot has been left to the human programmer of the robot. However, increasing robot and task complexity makes the design difficult. Consequently, automatic design process using artificial evolution is advocated to synthesis the controllers. Our work is concerned with exploiting the dynamics of local interactions between different behavioral levels of hierarchical modular structure for autonomous robot using genetic programming to achieve coherent goal-directed global behavior. Our proposal is exemplified and evaluated experimentally through a simulated mobile robot task. The robot autonomously navigates to a goal destination within an obstacle-ridden environment using its controller.
|