CogARCH: A Cognitive Agent Framework to Simulate Wayfinding by Architecture in Future Buildings

Findings from cognitive science link the architectural com- plexity of multilevel buildings with occupants’ difficulty in orienting and finding their way. Nevertheless, current ap- proaches to modelling occupants’ wayfinding reduce the rep- resentation of 3D multilevel buildings to isolated 2D graphs of each floor. These graphs do not take account of the interplay between agents’ 3D field of view and buildings’ 3D geometry, topology, or semantics, yet these are necessary to inform occu- pants’ path differentiation during wayfinding. Instead, agents are often modeled as unbounded and rational, able to calculate complete paths towards goals that are not immediately visible using direct routing algorithms. In turn, simulated behavior in most cases is unrealistically optimal (e.g. shortest or fastest route). This gap may hinder architects’ ability to foresee how their design decisions may result in suboptimal wayfinding be- havior, whether intended or not. To bridge this gap, the paper presents cogARCH, a computational, agent-based simulation framework. cogARCH is grounded in research on spatial cog- nition and heuristic decision making to support pre-occupancy evaluation of wayfinding in multilevel buildings. To demon- strate the relevance of cogARCH to architectural design, we apply it to assess wayfinding performance across three archi- tectural variations of a multilevel education building. Pre- liminary results showcase significant variability in cognitive agents’ wayfinding performance between building scenarios. In contrast, behavior of shortest-path agents sampled across respective conditions displayed significantly less variance and thus failed to reflect potential effects of architectural changes applied to 3D building configuration on wayfinding behavior.