Towards real-world capable spatial memory in the LIDA cognitive architecture

The ability to represent and utilize spatial information relevant to their goals is vital for intelligent agents. Doing so in the real world presents significant challenges, which have so far mostly been addressed by robotics approaches neglecting cognitive plausibility; whereas existing cognitive models mostly implement spatial abilities in simplistic environments, neglecting uncertainty and complexity. Here, we take a step towards computational software agents capable of forming spatial memories in realistic environments, based on the biologically inspired LIDA cognitive architecture. We identify and address challenges faced by agents operating with noisy sensors and actuators in a complex physical world, including near-optimal integration of spatial cues from different modalities for localization and mapping, correcting cognitive maps when revisiting locations, the structuring of complex maps for computational efficiency, and multi-goal route planning on hierarchical cognitive maps. We also describe computational mechanisms addressing these challenges based on LIDA, and demonstrate their functionality by replicating several psychological experiments.