RESEARCH RECORD / ForesightNav
* Equal contribution.
METHOD / FINDINGS
A training-free navigation framework that combines semantic maps, graph-based target verification, and geometric frontier exploration for object, instance, and text goals.
Context-aware semantic value maps guide broad search while online scene graphs verify fine-grained goal matches. Semantic score dispersion and graph-matching confidence choose among geometric exploration, semantic exploration, partial-match exploration, and target verification.
+18.6 percentage points in RoboTHOR success over UniGoal. RoboTHOR ObjectNav benchmark; percentage-point increase, not relative percent. (Table 1, p. 6.)
Geometrically verified task graphs and state-aligned reinforcement learning help navigation agents assess progress, recognize failure, and recover over long horizons.
A training-free navigation framework that combines semantic maps, graph-based target verification, and geometric frontier exploration for object, instance, and text goals.
