Several control and robotics problems can be addressed without the need for complex analytical solutions. This work investigates the use of genetic programming (GP) for some tasks, including the stabilization of pendulums of nonlinear dynamics and a mobile car-like robot navigation problem. As a simulation tool, a library focused on reinforcement learning is used, allowing a qualitative comparison between these methods. While deep reinforcement learning algorithms like DQN and DDPG are more sample-efficient, genetic programming offers a more interpretable alternative with similar raw computational time in some of the tasks.