I enjoyed the BoingBoing post Robots made from tree branches learns how to walk:
Made from servo motors and tree branches, these robots were trained to successfully walk across the floor. Unlike supervised machine learning, which trains models based on known-correct answers, in reinforcement learning, researchers train the model by having an agent interact with an environment. When the agent’s actions produce desired results, it gets positive feedback.
The research paper is available here: Improvised Robotic Design with Found Objects (pdf).