Abstract
The influential concept of cognitive maps envisions that the brain builds mental representations of objects, barriers, and goals. Computational models show how these representations guide goal-directed behavior, such as planning novel routes to maximize rewards. One key feature of flexible cognitive representations is compositionality, the ability to build complex structures by recombining simpler parts. However, how this applies to neural representations of cognitive maps and map-based planning remains unclear. Compositionality can be difficult to reconcile with efficient planning, as reusing components may limit efficiency. Here, we propose a novel computational model for efficiently creating and planning with compositional predictive maps, which successfully reproduces response fields in the medial entorhinal cortex, particularly object vector cells and grid cells. The model treats each object as an alteration to a baseline map linked to open space, creating predictive maps by combining object-related representations compositionally, providing insights into brain processes supporting efficient, flexible planning.