Skip to content

JPL Precision Landing: G-FOLD + ADAPT/FOALS → Mars 2020 Perseverance

Last updated: 2026-04-07 (Session 81)

Summary

Two FO projects from JPL's Robotics Section developed and flight-tested the Fuel Optimal Large Divert Guidance (G-FOLD) algorithm and Lander Vision System (LVS) terrain-relative navigation on Masten Space Systems' Xombie VTVL vehicle (2012–2014). The combined system — called ADAPT (Autonomous Descent and Ascent Powered-flight Testbed) — flew on NASA's Mars 2020 Perseverance rover (February 18, 2021), landing within 5 meters of target. This is the strongest confirmed mission infusion case in the FFRDC/academia tier: a FO parabolic/VTVL flight test directly enabled landing on Mars.

Outcome category: Mission Infusion (Mars 2020 Perseverance) — confirmed
Confidence: Confirmed (NASA and JPL acknowledge causal link explicitly)
Dollar tracked: >$2.5B (Mars 2020 mission cost; FO cost to G-FOLD/FOALS: ~$2M est.)

FO Projects

12272 — G-FOLD: Fuel Optimal Large Divert Guidance

  • Lead org: Jet Propulsion Laboratory (FFRDC/UARC), Pasadena CA
  • PI: Behcet Acikmese (JPL); Co-I: Andrew E. Johnson (JPL)
  • Period: 2012-09-30 to 2014-03-30
  • TRL: 3 → 5
  • Description: Developed onboard G-FOLD algorithm to compute fuel-optimal large-divert trajectories in real-time (<1 second) for powered descent. Customized for the Masten Xombie VTVL testbed.
  • FO flights: September 20, 2013 on Masten Xombie — simulated worst-case landing with real-time divert calculation to avoid hazard

91418 — FOALS: Fuel Optimal and Accurate Landing System

  • Lead org: Jet Propulsion Laboratory (FFRDC/UARC), Pasadena CA
  • PI: Andrew E. Johnson (JPL); Co-I: Joel Benito (JPL)
  • Period: 2014-08-27 to 2018-04-17
  • TRL: 4 → 6
  • Description: Integrated G-FOLD with terrain-relative navigation (LVS) in a ruggedized payload called ADAPT. Real-time closed-loop terrain detection + fuel-optimal divert on Masten Xombie.
  • FO flights: December 4 and December 9, 2014 — closed-loop LVS + G-FOLD integrated test; first real-time TRN+divert demonstration in a relevant environment

Library link confirms Perseverance infusion: FOALS project page links to "JPL's Terrain-Relative Navigation Technology Set to Launch on Mars 2020 Perseverance Rover" (parabolicarc.com, 2020) and "Terrain Relative Navigation: Landing Between the Hazards" (NASA Science).

Technology Chain

G-FOLD algorithm (UT Austin/JPL, ~2009–2012)
  → FO [12272]: ADAPT G-FOLD demo on Masten Xombie (Sep 2013) → TRL 3→5
  → FO [91418]: FOALS = G-FOLD + LVS integrated on Xombie (Dec 2014) → TRL 4→6
  → Mars 2020 LVS flight software (Perseverance EDL system)
  → February 18, 2021: Perseverance lands within 5m of target in Jezero Crater

What the LVS does on Mars 2020

During Perseverance's EDL: 1. Terrain-relative navigation (TRN): LVS fuses descent camera imagery with pre-loaded orbital maps to compute precise position relative to the landing site — without GPS. 2. Hazard detection: Identifies boulders, slopes, and craters in the landing zone. 3. G-FOLD divert: If the computed trajectory would land in a hazard, G-FOLD calculates a fuel-optimal divert path in <1 second. 4. Outcome: Perseverance landed within 5m of the target, in terrain that would have been inaccessible under Apollo-era precision.

PI Andrew Johnson developed the LVS for Mars 2020. The FO work on Xombie (ADAPT campaigns 2013–2014) was the critical real-world closed-loop test that gave NASA confidence to include TRN+divert guidance in the flight software.

Upstream Lineage

  • G-FOLD algorithm: Behcet Acikmese (JPL, later UT Austin) developed the convex optimization foundation, funded initially by NASA GCD/ARMD. The FO flight test was the first real-time onboard validation.
  • LVS: Andrew Johnson (JPL robotics) developed terrain-relative navigation. Pre-FO work was in Mars Science Laboratory (Curiosity) EDL analysis and JPL SLAM algorithms.
  • Masten Space Systems served as the VTVL flight testbed provider — not a developer of the algorithm.

Downstream Outcomes

Outcome Evidence Confidence
Mars 2020 LVS flight NASA/JPL explicitly credits FO ADAPT work (2013–2014) as validation enabling Mars 2020 inclusion Confirmed
Perseverance landing ±5m AIAA SciTech 2022 paper: "Mars 2020 Lander Vision System Flight Performance" (DOI 10.2514/6.2022-1214) Confirmed
SPLICE program G-FOLD + TRN became core technologies of NASA SPLICE (Safe and Precise Landing – Integrated Capabilities Evolution), now targeting future Artemis lunar landings Confirmed (NASA SPLICE program page)
Draper DMEN CP-12 SPLICE algorithms ported to Draper DMEN for CP-12 CLPS mission ($73M contract, launch ~2026) — see draper-precision-landing.md Suggestive

Note on dollar amounts: The Mars 2020 mission cost ~$2.5B. Attributing FO causality to a fraction of that is methodologically unsound. The appropriate claim is: FO provided the critical validation data that de-risked TRN+G-FOLD for Mars 2020 inclusion. The counterfactual (what would NASA have done without FO) is unknowable. Cost of FO projects [12272]+[91418] is estimated ~$2M total (FFRDC overhead + Masten flights).

Pattern Notes

FFRDC precision landing archetype: JPL used FO to conduct the one test that could not be replicated in a lab — real-time closed-loop GNC on a free-flying vehicle. Ground testing could validate individual algorithms; only a real flight could validate integrated TRN + divert guidance simultaneously. This is the ideal FO use case for an FFRDC: one targeted real-world validation that unlocks mission inclusion.

Contrast with NDL: Both are FFRDC/center-led FO projects → Mars/Moon mission infusion. NDL (Psionic, IM-1) went through commercial licensing; JPL TRN/G-FOLD went directly into a NASA mission. The institutional path is different, but both represent FO validating a technology that then flew on a planetary surface.

Session 52 Refresh (2026-04-07)

G-FOLD Heritage in SpaceX Landing

G-FOLD → SpaceX confirmed. PI Behcet Acikmese developed the lossless convexification framework at JPL that became G-FOLD. Lars Blackmore (SpaceX, responsible for Starship EDL) adapted the same convex optimization approach for all SpaceX rocket landings. The lossless convexification technique proven in FO flights on Masten Xombie (2013–2014) is now the algorithmic foundation for every Falcon 9 booster landing (~300+ successful landings as of 2026) and Starship development. This makes G-FOLD arguably the single highest-impact technology validated through Flight Opportunities — not by mission infusion count, but by total flights using the validated algorithm.

Acikmese career update: Acikmese left JPL and is now Professor, William E. Boeing Department of Aeronautics and Astronautics, University of Washington, Seattle. He gave an invited talk at Princeton (Feb 2026) on "Convex Optimization Based Optimal Control." His UW lab maintains the open-source SCPToolbox.jl for sequential convex programming trajectory optimization, which includes a Starship landing "flip" maneuver implementation.

SPLICE Continuation

SPLICE tested on New Shepard NS-13 (Oct 2020) and NS-17 (Aug 2021). The Safe and Precise Landing – Integrated Capabilities Evolution program continues as a GCD (Game Changing Development) project integrating TRN + NDL + Hazard Detection Lidar for lunar landing. SPLICE components may be included in Blue Origin's Blue Moon Pathfinder Mission 1 (NET early 2026), which targets 100m landing precision at the lunar south pole.

Blue Moon Pathfinder Mission 1 is a Blue Origin robotic lunar landing mission demonstrating the Blue Moon Mark 1 cargo lander, including BE-7 engine, precision landing, and cryogenic fluid systems. The precision landing requirement (100m) is modest compared to what SPLICE enables (<50m), suggesting SPLICE components may be available but not fully integrated for the pathfinder.

Assessment Update

The downstream impact of this FO arc is now significantly broader than originally assessed: 1. Mars 2020 Perseverance — confirmed mission infusion (TRN/LVS) 2. SpaceX Falcon 9/Starship — G-FOLD algorithm heritage in all booster landings 3. SPLICE lunar program — extending to Artemis and CLPS landers 4. Draper CP-12 — SPLICE algorithms ported for lunar far-side (see draper-precision-landing.md)

This is the broadest downstream tree in the FO portfolio: one FO flight test in 2013 → Mars landing (2021) + SpaceX operations (2015–present) + lunar landing program (2020s).

Session 81 Refresh (2026-04-07)

CP-12 Delayed to 2030

Major schedule slip. ispace announced the ULTRA lander redesign (March 2026), replacing the previous engine vendor with one that "has a proven track record of operation in past lunar missions." The CP-12/Draper mission (previously targeting ~2027) is now NET 2030 — pending NASA approval. ispace renumbered its missions: CP-12 is now "Mission 5" in ispace's revised sequence (Mission 3 in 2028, Mission 4 in 2029, Mission 5/CP-12 in 2030). This 3-year delay puts significant pressure on the Draper DMEN precision landing demonstration.

Impact on SPLICE/DMEN arc: The CP-12 delay means SPLICE-derived algorithms (Draper DMEN for lunar far-side) won't see flight validation until ~2030 — 17 years after the original FO G-FOLD test on Masten Xombie (2013). See draper-precision-landing.md.

Blue Moon Pathfinder Mission 1 Update

Blue Moon MK1 "Endurance" remains NET 2026 for launch on New Glenn from LC-36. Confirmed payload is SCALPSS (Stereo Cameras for Lunar Plume Surface Studies, $6.1M NASA). No SPLICE sensors confirmed in the payload manifest — the 100m precision landing target will use Blue Origin's own GNC, not NASA SPLICE. This corrects Session 52's speculation that SPLICE components "may be included."

Lars Blackmore — National Academy of Engineers

Lars Blackmore was elected to the National Academy of Engineers in 2026. This is explicit recognition of the lossless convexification work that connects directly to FO's G-FOLD validation. Blackmore is titled "Senior Principal Mars Landing Engineer" at SpaceX and is responsible for Starship EDL. SpaceX is planning for the first Starship landing back at the launch site in 2026 — a milestone that will use the convex optimization lineage validated through FO.

Acikmese Continued Publication

Acikmese published at AIAA SciTech 2025 (Orlando, Jan 2025): "Sequential Convex Programming for 6-DoF Powered Descent Guidance With Continuous-Time Compound State-Triggered Constraints" (arXiv 2510.09610). This extends the CT-SCvx method with compound state-triggered constraints for real-time PDG — the next generation of the algorithm lineage that started with G-FOLD. His UW lab continues to advance the algorithmic frontier that FO helped validate.

JPL Vibration Testing for Blue Ghost M2

JPL's vibration and acoustic testing facilities were used to test a full-scale model of Firefly's Blue Ghost Mission 2 spacecraft (Dec 2025), with lessons learned to be applied to upcoming CLPS missions. This is JPL infrastructure supporting the broader CLPS/precision landing ecosystem, though not directly G-FOLD/SPLICE related.

Updated Downstream Tree

FO G-FOLD on Masten Xombie (Sep 2013)
  → Mars 2020 Perseverance LVS (Feb 2021) — CONFIRMED
  → SpaceX Falcon 9/Starship booster landings (2015–present, 300+) — CONFIRMED heritage
    → Lars Blackmore elected NAE 2026
    → Starship launch-site landing planned 2026
  → SPLICE GCD program (NS-13 Oct 2020, NS-17 Aug 2021)
    → Draper DMEN CP-12 (now NET 2030 — delayed from ~2027)
    → Blue Moon Pathfinder (NET 2026 — NO SPLICE sensors confirmed)
  → Acikmese academic legacy: UW SCPToolbox.jl, AIAA SciTech 2025 paper

Assessment

Unchanged: broadest downstream tree in FO portfolio. The CP-12 delay to 2030 weakens the near-term SPLICE lunar demonstration path, but the core impact (Mars 2020 + SpaceX operations) is already locked in. Blackmore's NAE election validates the algorithmic contribution at the highest professional level. The G-FOLD → lossless convexification → SpaceX rocket landing pipeline remains the single highest-impact technology chain validated through Flight Opportunities.

Sources: NASA JPL news releases (2013, 2014, 2020); AIAA SciTech 2022 Mars 2020 LVS flight paper; JPL TRN website; TechPort library links for [91418]; NASA SPLICE program page; Lars Blackmore lossless convexification website (larsblackmore.com); UW ACL SCPToolbox.jl; Princeton MAE event (Feb 2026); Blue Moon Pathfinder Wikipedia; ispace ULTRA announcement (Mar 2026); SpaceX Starship landing plans; research.com Acikmese profile; AIAA SciTech 2025 paper (arXiv 2510.09610); JPL Blue Ghost M2 vibe testing (Dec 2025)