Our Blog

Turning a YOLO bounding box into a real-world ground coordinate — the pixel-to-ground pipeline, an interactive trial replay, and what our error tests revealed about camera tilt.

Repeated 50 m drop tests validated the passive two-stage delivery — clean release, controlled descent, no damage — clearing the way for a faster 35–15 m configuration.

Redesigning the beacon payload holder across three versions to protect the beacon while cutting bulk and weight and improving survivability.

Comparing our predicted battery usage against real flight measurements to sharpen mission planning and autonomous endurance estimates for ASCENT-1.

Slowing a 50 m drop with no motor: an FRT-G2-600 rotary damper on a 28 mm braked pulley brings the payload down to a safe hand-off speed.

Evolving the water-bottle payload holder from Version 1 to a lighter, better-aligned Version 3, now approved for active testing.

Bringing perception, communication, and flight together into one autonomous mission — confirm, localize, commit, and a LOITER orbit that refines the target to about two meters.

Matching PETG, ASA, ASA Aero, TPU, and carbon-fibre nylon to each part by the heat, weight, and load it has to survive in flight.

The mission logic that plans the search, debounces detections, and turns a camera pixel into GPS coordinates the drone flies to on its own.

A passive two-stage drop that replaces a motor-driven winch — one shared mechanism that delivers either the water bottle or the beacon.

Mapping every subsystem of ASCENT-1 onto one electrical architecture, from the Orange Cube+ and Jetson Orin Nano to the power distribution boards and payload winch.

A real-time object-detection pipeline on 15 watts — raw Bayer capture, TensorRT YOLO on a Jetson Orin Nano, and a live RTSP overlay to the safety pilot.

Collecting and annotating a custom aerial dataset of tents under competition-like conditions to improve the robustness of the team's YOLO perception pipeline.

Securing sponsorships from the Dream Realization Lab and TASNE3D — lab access, 3D printers, and discounted filament that accelerated the ASCENT-1 build.

Building a fault-tolerant MAVLink bridge between the Jetson and the Cube Orange+ — a companion-computer dead-man switch, a single-reader telemetry pump, and the field bugs that shaped both.

Standing up an in-house 3D-printing workflow on three Bambu Lab printers to prototype, iterate, and print ASCENT-1's structural parts overnight.

How the ASCENT-1 autonomy stack fits together — a Jetson for perception and mission logic, a Cube Orange+ for flight, one MAVLink link between them, and a state machine that always has a safe exit.

Selecting a lithium-ion battery configuration that balances endurance, weight, and the competition's easy-to-transport limit for ASCENT-1.

Designing a research platform to evaluate autonomous UAV search algorithms using the official SUAS 2026 mission environment before flight testing.