Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• OnFly improves task success rate in aerial vision-language navigation from 26.4% to 67.8% compared to the strongest state-of-the-art baseline.result 95%
• OnFly uses a shared-perception dual-agent architecture that decouples high-frequency target generation from low-frequency progress monitoring to stabilize decision-making.method 90%
• OnFly employs a hybrid keyframe-recent-frame memory to preserve global trajectory context while maintaining KV-cache prefix stability for reliable long-horizon monitoring.technical 85%
• OnFly uses a semantic-geometric verifier that refines VLM-predicted targets for instruction consistency and geometric safety using VLM features and depth cues.technical 90%
• OnFly's receding-horizon planner generates optimized collision-free trajectories under geometric safety constraints, improving both safety and efficiency.technical 85%
• OnFly enables fully onboard, real-time zero-shot aerial vision-language navigation.technical 90%
• The system might face challenges with environmental variability beyond tested conditions, such as unexpected weather or dynamic obstacles.limitation 80%
• Fully onboard real-world flights validate OnFly's feasibility for real-time deployment.result 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card