Multi-layered architectures for autonomous systems

• Autores: José Carlos González Dorado
• Directores de la Tesis: Fernando Fernández Rebollo (dir. tes.) , Angel García Olaya (codir. tes.)
• Lectura: En la Universidad Carlos III de Madrid ( España ) en 2020
• Idioma: español
• Tribunal Calificador de la Tesis: María Araceli Sanchís de Miguel (presid.) , José María Cañas Plaza (secret.) , Carlos Hernández Corbato (voc.)
• Enlaces
  • Tesis en acceso abierto en: e-Archivo
• Resumen

  • Deliberation is a key feature to endow advanced intelligence to autonomous agents. Complex use cases, specially robotic applications which require a considerable degree of autonomy, have to be coordinated by some kind of control architecture. However, reusing existing architectures can be difficult because use cases are very heterogeneous. The main goal of this thesis is to ease the use of standard deliberative techniques in use-case oriented cognitive architectures for autonomous systems. This thesis contributes with new declarative languages to model use cases, new control architectures, new software developments and new methodological guidelines of how to apply these concepts in engineering processes. To justify the contributions, this thesis presents the design of NAOTherapist, an autonomous social robot for upper-limb pediatric rehabilitation. Its control architecture is a prototype that maximizes its generalization capabilities. However, several limitations arose when reusing this architecture in other projects as the Clarc robot (geriatric assessment) and a CoBot robot (office delivery tasks). The discussion about NAOTherapist is used to highlight these limitations and determine how to overcome them. The main implementation after all these approaches is Mlaras (Multi-Layered ARchitecture for Autonomous Systems), which is a generic architecture to ease the development of autonomous intelligent systems. It also allows non-expert users to directly modify it to refine the use-case definition. Mlaras is focused on automated planning as the core of its deliberative processes, which are separated into several layers to take advantage of the hierarchical abstraction levels present in many use cases. An instance of Mlaras has been developed for the autonomous agents of a simulated logistics competition use case to further evaluate the capabilities of the new architecture. The experimentation concludes that the contributions help to fill the gap between the use case definitions and the actual developed applications.