City Walk: Embodied Locomotion Improves Route Efficiency and Spatial Memory in a Virtual City

Spatial navigation is a relevant target for assessment and training that relies on the interplay between egocentric route execution and allocentric, map-like representations. Virtual Reality (VR) enables ecologically plausible navigation tasks under experimental control, yet outcomes can strongly depend on immersion and locomotion interfaces that determine the availability of self-motion cues. In this paper, we present City Walk, a VR serious game designed to support implicit training and assessment of spatial navigation in a urban environment. The experience begins with guided acclimatization and gradually shifts to unguided exploration, time pressure, obstacle-induced re-planning, and ends with a map-based landmark placement task. City Walk implements two interaction modalities: Desktop VR (DVR) and Enhanced-Immersive VR (E-IVR), which combines a Head-Mounted Display (HMD) with an omnidirectional treadmill. We report a pilot between-subjects study comparing the two conditions. The protocol comprises five navigation levels with increasing demands, as well as a landmark placement test on an overhead map, supported by in-app logging and gaze-based landmark observation in the E-IVR build. E-IVR yielded substantially higher route efficiency and improved map-based landmark placement, while level completion times tended to be longer. User-centered questionnaires indicated comparable usability and tolerability across conditions.