← Back to work
MiniAug 2026 - Sep 2026Co-designer

Starship V3: Two-Stage Retropropulsive Landing Study

For the BIRST 2026 Level 3 challenge, we designed and simulated a two-stage vehicle in OpenRocket for a Falcon 9 Crew-9 mission profile. With no parachutes or active steering and only pre-timed burns, the simulation produced touchdown speeds of 14.9–16.7 m/s against a 5 m/s target and about 0.9 km of wind drift; this was a simulation, not a physical launch.

A focused simulation project with a clear question: how close can pre-timed burns alone get to a soft landing? Made for the BIRST 2026 Level 3 challenge.

OpenRocketOpenWind
+3 moreOpenEarth / Google EarthRetropropulsive landingWind modelling
Starship V3: Two-Stage Retropropulsive Landing Study preview

The question

The BIRST 2026 Level 3 (Falcon Rocket) task is to simulate a Falcon 9 Crew-9 launch from Cape Canaveral, reach a 3,000 m apogee and land both stages on their own engines. We designed and simulated a two-stage vehicle in OpenRocket. Real landings close a loop: sensors measure the descent and the vehicle adjusts continuously. We took the strictest version of the task and removed that loop, with no parachutes, no active stability control and only burns timed in advance, to ask what the loop is worth. I designed the simulation with one teammate.

What we found

The OpenRocket simulation reached 2,999.97 m against the 3,000.00 m target and peaked at Mach 0.973, so it stayed subsonic. Simulated landing missed the 5.0 m/s speed specification: the upper stage reached 16.7 m/s and the booster 14.9 m/s, roughly three times the target, because fixed-delay ignition over-corrected the vertical momentum. Uncompensated crosswinds produced 0.86 to 0.95 km of simulated downwind drift.

Why fixed timing falls short

We traced the touchdown miss to the ignition delays. They are chosen before launch, so no burn can respond to how the descent actually unfolds, and any small difference between the planned and the real fall goes straight into the touchdown speed. The wind result has the same cause: nothing in the vehicle senses the crosswind or actively steers against it, so the wind sets the drift.

How it was built

The OpenRocket simulation model used a Loki K627LR ascent-motor specification (1,526 N·s). The Cape Canaveral (SLC-40) site was recreated in OpenWind with a 30-layer extended ISA wind profile (crosswinds of 3.1 to 14.4 m/s), and simulated trajectories were viewed in Google Earth through OpenEarth. The model used AeroTech HP-H195NT and Cesaroni 168H410-14A motor specifications for the upper stage and booster, with fixed ignition delays of 51.5 s and 65.0 s and custom swept fins standing in for Falcon 9's grid fins. These are simulation inputs, not evidence of a physical launch.

What the miss is worth

The simulated miss provides a baseline for examining the effects of fixed-delay burns and uncorrected crosswind: about three times the target touchdown speed and nearly a kilometre of drift. It is not a physical flight result.

What we would try next

Better burn timing, different motor choices and launch-angle compensation for wind are the changes we would try next.