<p>Social learning in robotic swarms enables robots to improve their behavior by exchanging or imitating the behavioral knowledge of their peers. This study focuses on embodied evolution, a form of social learning in which an evolutionary algorithm is distributed across a population of robots, allowing each robot to evolve its controllers onboard through local interactions. This study proposes a simple embodied evolution approach that enables robots to evolve both the topology and weights of their neural network controllers during operation. The proposed approach utilizes a topology and weight neuroevolution method that uses only mutations for genetic variation. The robot controllers are evolved using an embodied evolution framework in a two-target navigation task conducted in computer simulations. The experimental results in a simulated task environment demonstrate that the proposed approach exhibits better performance than conventional fixed-topology controllers.</p>