Implementation in ROS2 of a Purposeful, Intrinsically Motivated Cognitive Architecture for Autonomous Robotics

Abstract

In the PILLAR-Robots project, a cognitive architecture has been designed to provide robots with Lifelong Open-ended Learning Autonomy (LOLA), which can be modulated to align with the requirements of the user across multiple domains. This architecture provides key functionalities such as perception, attention, self-motivation, and decision-making. For practical applications in robotics, a software implementation based on the Robotic Operating System (ROS2) has been developed to ensure ease of integration with robots, modularity, and interoperability. To achieve this, the system separates ROS2 execution from high-level cognitive processes via a specialized middleware. The capability of the architecture to complete tasks has been demonstrated by initial results in simulated environments.