Performance of Navigator Module: UAV Vision-Based Navigation Simulation Results

Gobinda Chandra Sarker published this dataset on figshare in April 2026. It contains simulation results for a novel vision-based Unmanned Aerial Vehicle (UAV) navigation system that integrates depthmap estimation with a Vision-Language Model for obstacle avoidance and path planning. The system was validated in AirSim using Blocks and the Downtown West environment, achieving a reported task completion rate of 0.98.

Use Cases

• Benchmarking vision-language models for autonomous navigation based on the described integration of depth maps and VLM output.
• Analyzing the performance of depth estimation models in dynamic environments based on the use of DepthAnything-V2.
• Comparing obstacle avoidance strategies in simulated urban and structured environments based on tests in the Blocks and Downtown West AirSim worlds.
• Evaluating cost-efficient navigation systems that forgo LiDAR and pre-existing maps based on the described system architecture.

Strengths

• Dataset is licensed under CC-BY-4.0, permitting broad reuse.
• The described system achieved a near-perfect task completion rate of 0.98 in simulations.
• File size is 5.5 KB, indicating a compact and likely focused dataset.

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• Data is limited to simulation results from specific AirSim environments, which may not fully represent real-world conditions.

Provenance

Source

Gobinda Chandra Sarker via figshare

Collection Method

Simulation results generated in AirSim using Blocks and Downtown West environments.

Time Range

Simulation date range is not specified.

Freshness

Last updated 2026-04-03 17:54:46; freshness should be verified.

Geography

Simulated environments (Blocks, Downtown West); no real-world geography.