Content area
This paper shows how genetic programming (an area under the umbrella of evolutionary computation) can be applied in two out of the six RL 2009 benchmark problems, such as the Acrobot and the Generalised Helicopter Hovering. The paper is organised as follows: the next section provides a short introduction to the field of genetic programming and how it can be seen as an automatic generator of controllers. Section 3 describes the acrobot problem and the experiments and results obtained by using a genetic programming approach. Section 4 describes the generalised helicopter hovering problem and the simulation results obtained using a controller evolved by genetic programming. The same section compares the genetic programming controller with the winner of the 2008 reinforcement learning competition. Finally, Section 5 summarises this work and outlines future desirable research.