Urbano,Paulo Jorge Cunha Vaz DiasSantos,Filipe Gomes Marques dos2026-01-192026-01-192025http://hdl.handle.net/10400.5/116696Tese de Mestrado, Informática, 2025, Universidade de Lisboa, Faculdade de CiênciasCommunication is a fundamental mechanism for information exchange, playing a central role in problem solving and closely intertwined with evolutionary processes. In artificial intelligence, particularly within evolutionary computing and evolutionary robotics, communication has been shown to significantly enhance the performance of evolved models in tasks that require coordinated behaviour among multiple agents. Novelty-based evaluation is a relatively recent evolutionary metric which, by replacing traditional fitness functions, can mitigate premature convergence to local optima and reduce evolutionary stagnation in complex problem domains. In this work, we propose that incorporating both communication and novelty-driven methods into neuro-evolutionary models can substantially improve their performance, thereby contributing to advances in evolutionary algorithms and evolutionary robotics. To support this hypothesis, we analyse the impact of these approaches in multi-agent systems using the well-known predator–prey capture problem with social robots in one-dimensional and two-dimensional environments.application/pdfporNeural EvolutionReinforced LearningArtificial Neural NetworksNoveltyCommunicationmaster thesis204176093