TravelAgent explores generative agents in the built environment. The project models pedestrian navigation, wayfinding, and free exploration across varied indoor and outdoor scenes, using multimodal observations and agent memory to simulate human-like decisions. The work positions generative agents as a method for testing urban scenarios, reading spatial experience, and supporting early design decisions.
TravelAgent system schemeWatch TravelAgent podcastAI Agents and Urban PlanningPath visualization in the 3D experiment environmentAgent paths and decision points across scenariosTerm frequency analysis of agent planning streamsSentiment analysis across agent paths
