Adding a Cognitive Architecture to ROS2 for Interactive Task Learning in TurtleBots

Abstract

This work presents an empirical study of an integrated cognitive model that combines ROS2 actions and sensory data with a cognitive architecture pyACT-R for a turtlebot4 robot. The robot operates in a grid-based environment containing obstacles, while initially not knowing their properties (unknown obstacles). Path planning is performed using the A* algorithm, and encountered unknown obstacles are classified through physical interaction: non-traversable obstacles are classified as solid and avoided in subsequent path planning, while obstacles that turn out to be traversable despite prior evidence are classified as passable. Based on repeated collision experiences, the robot adapts its navigation strategy by switching from shortest-path strategy to a risk-aware one that avoids all obstacles. Experimental results demonstrate that the cognitive extension enables adaptive strategy selection and memory-based learning of obstacle properties. This highlights the potential of cognitive approaches for interactive task learning in mobile robots.