Dual-Horizon model improves drone navigation with foresight and memory
ForeFly: A Dual-Horizon World Action Model for Aerial Vision-Language Navigation
Computer Vision and Pattern Recognition
Summary
Following directions for drones flying in 3D spaces is hard because they need to understand both nearby and faraway things. The authors create ForeFly, a model that looks ahead in two ways: shortly ahead to maintain smooth movement and further ahead to follow the whole route correctly. It uses memories of what it recently saw and important route details to help it decide where to go next. This approach works better than others on standard drone navigation tests.
What this means in practice
- •For drone operators: Enhance drone guidance systems for longer, complex flights by using foresight models that predict both near and far future path conditions.
- •For robotics engineers: Design autonomous UAVs that improve path-planning and navigation accuracy by integrating dual-horizon visual foresight models.
Authors
Kunhui Wang, Xintong Zhang, Junyu Gao, Changsheng Xu
Abstract
Aerial Vision-Language Navigation (AVLN) requires UAVs to maintain reliable instruction following over long trajectories in complex 3D environments. However, existing AVLN approaches are predominantly reactive or limited to single-horizon prediction, overlooking complementary future cues across different temporal horizons. To address this limitation, we propose ForeFly, a dual-horizon latent world action model that predicts both a proximal future for local continuity and an adaptive route-critical future for long-range guidance. Horizon-specific foresight queries are primed with recent and route-critical visual memories, providing history-aware context for future prediction. To exploit their distinct roles in action generation, we introduce Foresight-Guided Action Refinement (FGAR), which asymmetrically exploits proximal foresight for local action enhancement and route-critical foresight for feature-wise correction and route-level guidance. Experiments on the TravelUAV and UAV-ON benchmarks show that ForeFly consistently outperforms strong baselines across seen and unseen settings, validating the effectiveness of dual-horizon foresight and FGAR learning. The code is available at: https://github.com/kunhuiW/ForeFly