UT-SARA-GQ: Off-Road Aerial-Ground Robot Localization Dataset

UT-SARA-GQ is an off-road aerial-ground localization dataset collected with a Clearpath Warthog robot at the University of Texas SARA test site. The processed benchmark contains 15 trajectories totaling 8.3 km and approximately 60,000 frames, split into 9 training, 2 validation, and 4 test trajectories. It was created by Dongmyeong Lee and released via the Texas Data Repository Harvested Dataverse, last updated on 2026-04-27.

Use Cases

• Train and evaluate aerial-ground localization models based on rectified ground observations and aerial orthophoto maps.
• Research cross-view matching and visual localization based on stereo images and depth data.
• Benchmark map alignment and odometry comparison algorithms based on synchronized GPS, odometry, and pose information.
• Study off-road robot navigation in environments with tree canopy, shadows, and visually repetitive terrain.

Strengths

• Provides 15 processed trajectories totaling 8.3 km of off-road robot data.
• Includes approximately 60,000 frames with rectified stereo images, depth, GPS, odometry, and pose data.
• Offers both raw ROS 2 bag recordings and processed, benchmark-ready sequences for flexibility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic and environmental bias inherent to the single test site.

Provenance

Source

University of Texas, collected at the UT SARA test site Grace Quarters environment.

Collection Method

Collected with a Clearpath Warthog robot, includes raw ROS 2 bag recordings and processed sequences.

Freshness

Last updated 2026-04-27 04:28:00; freshness should be verified.

Geography

Outdoor field site at the UT SARA test site (Grace Quarters environment), UTM zone 18N.