# BepiColombo and Psyche: mission references for LOTUS examples

Research reference date: **10 November 2026**. Primary mission publications and live ESA/JPL ephemerides were checked independently of the older repository audits. This report prioritizes trajectory reconstruction; subsystem information is a reference inventory for subsequent work, not a claim that LOTUS reproduces flight engineering or telemetry.

The current examples use actual LOTUS plugin HTTP solves and independent NASA/JPL comparisons. Both primary candidates fail physical screening under the declared nominal hardware assumptions; neither establishes a successful flight reproduction. See the [plugin validation methodology](plugin-validation-methodology.md). The earlier 174-day BepiColombo private-function design and 112-day Psyche coast probe are archived experiments, superseded as public validation evidence.

## Findings that control the implementation

1. BepiColombo's operational route is **Earth launch → Earth → Venus → Venus → Mercury × 6 → Mercury capture**. The existing STAR optimization benchmark has different launch dates, encounter sequence and propulsion. Retain its identity as a benchmark.
2. Both missions depend on distributed solar electric propulsion. Joining encounters with ballistic Lambert arcs cannot reproduce their flown trajectories. Display/import sampled operational ephemerides as reference trajectories, and distinguish that from a newly optimized solution.
3. Current Horizons BepiColombo coverage reaches April 2027; current Psyche provider coverage stops in **February 2029**, before asteroid arrival. Independent checks reveal a large Psyche prediction discontinuity on **1 March 2027**, so the coherent LOTUS reference ends before that transition. Do not silently extend either reference to planned arrival using an old design kernel.
4. Numerical interpolation error, source orbit-determination uncertainty, and agreement with rounded press-release altitudes are different quantities. Public covariance was not established here; no flight-position accuracy claim follows from a smooth line.

The underlying products are [encounter-reference.json](encounter-reference.json), ten preserved raw encounter responses listed in the [source manifest](source-manifest.json), and [a reproducible encounter calculation](compute_encounter_references.py). Each JSON encounter includes its exact API URL, state at the nearest one-second sample, explicit altitude radius, and recommended sampling windows.

## Ephemeris identity, provenance and conventions

| Item | BepiColombo | Psyche |
|---|---|---|
| Horizons spacecraft target | `-121` | `-255` |
| Object header revision retrieved | 3 November 2026 | 3 November 2026 |
| Fit/data cutoff stated by provider | 22 October 2026 | 27 October 2026 |
| Later coverage | Prediction | Prediction |
| Current listed coverage, TDB calendar | 20 Oct 2018 02:14 → 10 Apr 2027 23:59 | 13 Oct 2023 15:21 → 24 Feb 2029 |
| Source | ESA `bc_mpo_fcp_00239_20181020_20270411_v02` | JPL reconstructed/operational sequence, then `psyche_ref_261102-290224_261007_v1` |

These are statements from preserved [BepiColombo](sources/bepicolombo-horizons-object.txt) and [Psyche](sources/psyche-horizons-object.txt) API object headers, not inferred from file dates. Public [Horizons](https://ssd.jpl.nasa.gov/horizons/manual.html#spacecraft-trajectories-in-horizons) serves spacecraft trajectories supplied by project navigation teams. The current source names and output headers must travel with generated data.

The ESA [operational metakernel, v434 dated 10 November 2026](https://spiftp.esac.esa.int/data/SPICE/BEPICOLOMBO/kernels/mk/bc_ops_v434_20261110_001.tm), now lists a newer `bc_mpo_fcp_00240_20181020_20270411_v02.bsp`. It is a different release from the Horizons copy. Preserve which release was used; do not imply a Horizons export used this newer ESA kernel. The metakernel also supplies DE432s, leapseconds, frames, structure, attitude and instrument kernels. Its complete [snapshot](sources/bc_ops_v434_20261110_001.tm) is retained.

The [NAIF Psyche archive](https://naif.jpl.nasa.gov/pub/naif/PSYCHE/kernels/spk/) distinguishes `rec` and `ref` kernels. The Mars flyby reconstruction `psyche_rec_260501-260713_260804_v1.bsp` covers **1 May 2026 23:59:59 UTC–13 July 2026 22:40:00 UTC**, identifies body `-255`, and names DE440/MAR097. Its [label](sources/psyche_rec_260501-260713_260804_v1.bsp.lbl) is retained. Current Horizons encounter output instead names **MAR099** for Mars: a standalone-kernel comparison must use matching planetary ephemerides before attributing differences to interpolation.

An older [merged Psyche reference label](sources/psyche_ref_231013-311101_231016_v1_merged.bsp.lbl), created October 2023, spans through 1 November 2031 and contains spacecraft `-255` plus project target asteroid `2000016`. This is useful for a separately labelled historical design scenario, not a current reconstruction beyond available Horizons coverage. In Horizons, `2000016` selects the flight-project target; `16;` selects the latest independently maintained asteroid solution. These are distinct from the spacecraft and can differ.

For repeatable exports use `EPHEM_TYPE=VECTORS`, `VEC_CORR=NONE`, `REF_SYSTEM=ICRF`, `REF_PLANE=ECLIPTIC`, `OUT_UNITS=KM-S`, `VEC_TABLE=3`, and an explicit center. `500@10` is heliocentric; `500@399`, `299`, `199`, and `499` identify Earth, Venus, Mercury, and Mars centers respectively. Do not subtract a barycenter from a body-centered state. [JPL API documentation](https://ssd-api.jpl.nasa.gov/doc/horizons.html).

Our independent encounter queries explicitly request `TIME_TYPE=UT`; their headers confirm modern dates are **UTC**, and their Julian dates are JDUT. Default vector epochs are TDB and must never simply be labelled `Z`. Continuous integration/interpolation should use one declared dynamical time scale; retain UTC conversion for human presentation. Future UTC conversion assumes the last known leap second and must be labelled accordingly. Geometric states have no light-time or stellar-aberration correction. [Horizons time and frame conventions](https://ssd.jpl.nasa.gov/horizons/manual.html).

## BepiColombo trajectory and chronology

Launch was 20 October 2018, 01:45 UTC, Ariane 5 from Kourou. ESA and JAXA carry complementary polar orbiters MPO and Mio, with MTM providing cruise propulsion and MOSIF shielding Mio. [ESA factsheet](https://www.esa.int/articles/about-us/directorates/science/bepicolombo/bepicolombo-factsheet).

The table below is **independently calculated from the current Horizons geometric ephemeris**, using a one-second bracket of relative radial velocity changing sign. Times are rounded to the nearest second; ranges are to the body center. The stored subsecond roots are numerical interpolation results, not subsecond flight-event certification.

| Encounter | Current ephemeris closest approach, UTC | Center range, km | Published surface altitude, km |
|---|---|---:|---:|
| Earth | 10 Apr 2020 04:24:57 | 19064.086 | about 12700 |
| Venus 1 | 15 Oct 2020 03:58:32 | 16773.730 | 10721.6 |
| Venus 2 | 10 Aug 2021 13:51:54 | 6604.375 | 552 |
| Mercury 1 | 1 Oct 2021 23:34:42 | 2639.324 | 199 |
| Mercury 2 | 23 Jun 2022 09:44:22 | 2638.590 | about 200 |
| Mercury 3 | 19 Jun 2023 19:34:25 | 2675.332 | 236 |
| Mercury 4 | 4 Sep 2024 21:48:49 | 2603.859 | about 165 |
| Mercury 5 | 1 Dec 2024 14:23:41 | 40066.128 | 37626 |
| Mercury 6 | 8 Jan 2025 05:58:52 | 2735.535 | about 295 |

Published event cross-checks: [Earth result](https://www.esa.int/media/bepicolombo-s-final-glimpses-of-earth), [Venus 1 navigation update](https://www.esa.int/Science_Exploration/Space_Science/BepiColombo/BepiColombo_flies_by_Venus_en_route_to_Mercury), [Venus 2 imagery](https://www.esa.int/ESA_Multimedia/Images/2021/08/Flying_by_Venus), [Mercury 1](https://www.esa.int/ESA_Multimedia/Images/2021/10/BepiColombo_meets_Mercury_annotated), [Mercury 2](https://www.esa.int/ESA_Multimedia/Images/2022/06/BepiColombo_s_second_Mercury_flyby), [Mercury 3 science analysis](https://doi.org/10.1051/0004-6361/202553870), [Mercury 4 result](https://www.esa.int/Science_Exploration/Space_Science/BepiColombo/BepiColombo_s_best_images_yet_highlight_fourth_Mercury_flyby), [Mercury 5 result](https://www.esa.int/articles/about-us/directorates/science/bepicolombo/bepicolombo-reveals-mercury-in-a-new-light), [Mercury 6 result](https://www.esa.int/media/bepi-mio-and-mtm-complete-their-final-mercury-flyby).

The live factsheet/Horizons prose still includes an **11 August** Venus-2 date; the dedicated flown-event record and ephemeris establish **10 August**. Horizons prose also retains a 20 June Mercury-3 date; actual encounter was **19 June**. Event press times and updated ephemeris minima differ by fractions of seconds or seconds. Surface heights depend on shape convention and source release. The JSON explicitly records spherical radii and does not equate derived spherical altitude with ellipsoid/geodetic height.

Reduced SEP power required replanning in 2024 and moved arrival from December 2025 to November 2026. The fourth flyby was adjusted to support that trajectory. [ESA replan announcement, 2 September 2024](https://www.esa.int/Science_Exploration/Space_Science/BepiColombo/Fourth_Mercury_flyby_begins_BepiColombo_s_new_trajectory).

SEP finished on **15 June 2026 at 15:24 CEST = 13:24 UTC**, according to ESA's 24 June update. After MTM release, MPO chemical propulsion supports arrival. [ESA SEP completion](https://www.esa.int/articles/about-us/directorates/operations/end-of-the-blue-glow-bepicolombo-turns-off-solar-electric-propulsion-for-mercury-arrival).

MTM separation succeeded **3 September 2026**, with expected onboard time 14:00 CEST = 12:00 UTC. A subsequent ESA article reports signal acquisition 15:49 CEST, while the live log says 15:52. Neither ground receipt time should be used as a separation state epoch. [ESA separation account](https://www.esa.int/articles/about-us/directorates/operations/bepicolombo-begins-mercury-arrival-with-mtm-separation-success), [live arrival log](https://www.esa.int/articles/about-us/directorates/science/bepicolombo/latest-updates-bepicolombo-s-arrival-at-mercury).

As of this report, **21 November 2026 capture, 9–10 December Mio separation and April 2027 science are plans**. MPO's final orbit is targeted for March 2027. Science-orbit design altitudes are MPO **480 × 1500 km**, Mio **590 × 11640 km**, both polar; these are not initial capture ellipses. [ESA arrival schedule](https://www.esa.int/articles/about-us/directorates/science/bepicolombo/latest-updates-bepicolombo-s-arrival-at-mercury), [ESA engineering mission paper](https://electricrocket.org/2019/824.pdf).

## Psyche trajectory and chronology

Psyche launched **13 October 2023 14:19:43 UTC** on Falcon Heavy from Kennedy LC-39A; spacecraft separation was about 15:21 UTC. Preserve this distinction when the spacecraft ephemeris begins after launch. [JPL spacecraft object record](sources/psyche-horizons-object.txt).

The only planetary gravity assist is Mars, **15 May 2026**. Our independent current reconstructed-ephemeris calculation yields **19:28:02.786 UTC**, **8000.219798 km from Mars center**, with a one-second root bracket. NASA's flown-event report gives **4609 km above the surface**, approximately **1000 mph heliocentric speed gain** and **1° orbital-plane adjustment**. The stale nominal Horizons description gives 4618 km; use its numerical reconstruction, not that nominal prose. [NASA result, 19 May 2026](https://www.nasa.gov/missions/psyche-mission/nasas-psyche-mission-aces-mars-flyby-targets-metal-rich-asteroid/).

On **28 September 2026** nearly continuous SEP began. The 22 October JPL update says approximately **90% duty** to summer 2029, interrupted for communications/engineering, single active Hall thruster, up to **250 mN**; roughly one-third of the launch xenon had been consumed. Power will decline with solar distance. [JPL October cruise update](https://www.jpl.nasa.gov/news/pedal-to-the-metal-nasas-psyche-heads-toward-asteroid-belt/).

The launch-era plan calls for approach in May 2029, gravitational capture in late July, and science starting August. Planned sequence is **A → B1 → D → C → B2**, with approximate altitudes **709, 303, 75, 190, 303 km**. B is split for illumination; low D has a different inclination. Orbit transfers take weeks/months and depend on the asteroid's unknown detailed gravity/shape. Dates are subject to revision. The prime plan ends November 2031. These are descriptive planning references only until a current complete kernel is available. [JPL mission plan](https://www.jpl.nasa.gov/press-kits/psyche/mission/).

## Encounter and leg windows for software

These are engineering recommendations derived for LOTUS visualization/validation, not mission operations windows. Each encounter in the JSON has explicit start/stop epochs computed from its own CA time.

| Purpose | Span around each CA | Initial maximum sample step |
|---|---|---:|
| Planet-relative context | ±3 days | 300 s |
| Approach/departure detail | ±6 hours | 60 s |
| Close flyby | ±30 minutes | 5 s |
| Independent CA validation | ±2 minutes | 1 s |

Use each pair of consecutive encounters as a heliocentric leg, retaining all intermediate revolutions. Start cruise sampling at 6 hours to 1 day with position **and velocity**; refine adaptively with independent midpoint queries. Use body-relative vectors for flyby detail to avoid subtracting two large interpolated heliocentric positions. Keep acceleration/discontinuity boundaries and ephemeris gaps explicit. Near launch, capture and orbit lowering, daily samples are inadequate.

For BepiColombo arrival distinguish approach/MTM release, capture, Mio delivery/release and MPO lowering. An arrival-day label is not a closest-approach or burn epoch. Use 3 September–21 November 2026 as approach context, 20–23 November as capture context, December as deployment context, and January–March 2027 for MPO lowering, only within the chosen kernel's coverage. Do not manufacture precise capture/burn timestamps from dates alone.

For Psyche, define launch→Mars and Mars→**the coherent cutoff before 1 March 2027 12:00 UTC** as reference legs. The provider's later February 2029 endpoint cannot be used across the independently verified discontinuity. Show planned arrival separately. A newly optimized Earth→Mars→Psyche scenario may use tightly bounded dates for the first two events; it still requires finite low-thrust optimization and target ephemeris matching. Setting exact flyby dates alone does not establish a close trajectory match.

For planner input, **fix** the reference launch and encounter epochs during replay. If the user explicitly creates a nearby optimization study, a useful initial search constraint is ±1 day around each historical encounter, with positive leg-duration bounds derived from consecutive epochs; that ±1 day is our search choice, not a published navigational tolerance. Use narrow epoch windows before broadening a search. Preserve the full encounter sequence and signed incoming/outgoing vectors, not merely target-body names. Derive revolution count and branch from the sampled ephemeris's unwrapped heliocentric longitude, rather than assigning every leg a zero-revolution Lambert branch. Time windows and an extra DSM cannot substitute for a distributed SEP control history; no ballistic solution should be labelled a reproduction solely because its event dates or total duration match.

## Subsystem inventory for the next phase

### BepiColombo

The mission paper gives a consistent **nominal** mass model: MPO dry **1146.6 kg**, MTM dry **1090.2 kg**, Mio/interface **274.7 kg**, MOSIF **120.8 kg**; dry sum **2632.3 kg**, nominal propellant **1410.9 kg**, total **4043.2 kg**. ESA's public ~4100 kg figure is rounded and uses a different presentation. Do not combine unlike inventories as measured mass. MPO and Mio instruments are included in spacecraft dry mass. [Mission overview, Table 1](https://link.springer.com/article/10.1007/s11214-021-00861-4).

MTM has four T6 gridded ion engines; the published mission envelope is **75–125 mN each**, up to **250 mN with two operating**. The 2019 design total-impulse requirement is **17.6 MN·s**. Planned navigation coasts were 30 days before and 7 after flybys; these are design constraints, not proof of actual thrust switching for every encounter. [ESA SEP engineering paper](https://electricrocket.org/2019/824.pdf). The independently documented as-loaded xenon quantity is **581 kg**, loaded June 2018. [ESA loading record](https://www.esa.int/media/loading-the-mtm-with-xenon).

MPO uses chemical propulsion for acquisition, with **22 N bipropellant** and **5 N hydrazine** thrusters. Mio is spin stabilized (~15 rpm), with attitude-control nitrogen rather than an interplanetary transfer engine. MTM solar area is about **42 m²**, using tilt to limit temperature. Mercury sunlight spans roughly **6290–14500 W/m²**. MPO uses heat pipes, high-temperature insulation and a protected radiator; Mio uses spin and its cruise sunshield. These are hardware/design references, not correlated thermal-network inputs. [Mission overview](https://link.springer.com/article/10.1007/s11214-021-00861-4).

MPO has a **1 m steerable HGA**, **35 W X/Ka transmitters**. Published nominal frequencies include X downlink **8.419 GHz**, X uplink **7.166 GHz**, telemetry Ka downlink **31.992 GHz**. MORE's separate Ka transponder is **1.9 W**, with **32.075/34.357 GHz** down/up. Malargüe's 35 m station and DSN Goldstone support radio science. Keep ranging/Doppler calibration separate from telemetry throughput. [MORE engineering paper](https://link.springer.com/article/10.1007/s11214-021-00800-3).

### Psyche

Launch-era public requirements state wet mass **up to 2747 kg including DSOC**, **21 kW at departure**, **2.3–3.4 kW at asteroid orbit**, four Hall thrusters with one active, and **up to 1085 kg xenon**. These are specification bounds, not exact current mass or power telemetry; the 2026 update rounds xenon to 1090 kg. [JPL quick facts](https://www.jpl.nasa.gov/press-kits/psyche/quick-facts/).

Psyche uses a modified Maxar 1300 chassis, JPL avionics, four gimballed **SPT-140 Hall thrusters**, reaction wheels and nitrogen cold gas. It has **no chemical combustion propulsion**. Three nitrogen tanks feed redundant strings; cold-gas thrusters are rated **0.45 N**, **70 s Isp**, with maximum **30% duty**. X-band telecom uses a **2 m fixed HGA**, three LGAs and **100 W TWTA**; published capabilities include up to **2 kbps uplink** and **180 kbps downlink** with the specified geometry/DSN support. These are design capabilities, not constant rates. [2025 spacecraft engineering paper](https://link.springer.com/article/10.1007/s11214-025-01222-1).

Thermal hardware includes insulation, embedded heat pipes, heaters, radiator louvers and about **300 temperature sensors**. Battery capacity is **144 Ah**, 13 cells. Launch-era dry mass is **up to 1616 kg**. Do not infer a chemical propellant pool by subtracting xenon from mass bounds. [JPL spacecraft description](https://www.jpl.nasa.gov/press-kits/psyche/mission/spacecraft/).

DSOC is a separate optical technology demonstration: **22 cm flight aperture**, **4 W flight laser**, **5 kW Table Mountain uplink**, **5.1 m Hale Telescope receiver**. It was not the baseline operational mission-data link; a radio link budget cannot model it by changing antenna diameter. [JPL DSOC design description](https://www.jpl.nasa.gov/press-kits/psyche/dsoc/).

## Validation, uncertainty and remaining inputs

The independent CA checks query each spacecraft directly relative to the actual planetary center, solve the range-rate sign change on a one-second grid, and preserve complete responses. Validate imported encounters against these center ranges and epochs. Position samples in the JSON belong to the **nearest grid epoch**, not the interpolated subsecond root. Spherical heights are explicitly derived and carry no geodetic claim.

An independent provider check also found a **57.917 km position discontinuity** in the current BepiColombo Horizons source, bracketed between **3 May 2023 21:33:46.696 and 21:33:46.697 UTC**. Uniform one-second output first exposed it; individually quoted calendar queries at 0.1 s, 0.01 s and 0.001 s isolated it, and both Sun-centered and Mercury-centered queries reproduced the jump. This is a discontinuity in the supplied reference states, not evidence of a physical maneuver. Interpolation must stop across such boundaries and disclose bounded unavailable intervals; decreasing the interpolation step cannot make a finite jump continuous. Evidence: [uniform output](sources/bepi-20230503-discontinuity.txt), [Sun-relative 0.1 s query](sources/bepi-20230503-discontinuity-10-0.1sec.txt), [Mercury-relative 0.1 s query](sources/bepi-20230503-discontinuity-199-0.1sec.txt), and [1 ms bracket](sources/bepi-20230503-discontinuity-10-0.001sec.txt). [Source snapshot checksums](source-manifest.json) cover all retained source files.

Psyche has a much larger prediction transition: a direct calendar query changes by **1,968,181 km between 1 March 2027 12:00:00 and 12:00:01 UTC**, although the earlier state velocity is only about **19.23 km/s**. This independently verifies that the current merged provider data cannot be treated as one continuous trajectory across this transition. A coherent example should end before that transition, rather than draw a connection or claim the remainder reaches the planned asteroid orbit. The provider's February 2029 overall coverage endpoint does not establish continuity. [Independent Psyche transition query](sources/psyche-20270301-discontinuity.txt).

For rendered interpolation, evaluate held-out provider states at interval midpoints and compare Euclidean position and velocity errors in the declared frame. Record the number of checks, maximum/RMS errors and worst interval. Tighten sampling until the declared tolerance passes; a pixel-perfect full-system view does not establish encounter fidelity. Test time conversion against known UTC event epochs and verify the importer rejects data outside coverage, duplicate/unsorted timestamps and non-finite states.

Preserve actual requested epochs with sufficient precision during import. A review of the first import found that parsing Horizons' displayed nine-decimal JDUT field introduced tens of microseconds of epoch quantization; on ten-second intervals this produced artificial velocity interpolation errors. The corrected importer recovers each exact submitted twelve-decimal JD token from its request, subtracts JD2451545 and multiplies by 86400 using decimal arithmetic before conversion to floating point. This also retains the original submitted time's small offset from an intended round second. Calendar `TLIST_TYPE=CAL` requests are another option for future imports. Additional same-epoch vector checks at encounters should supplement midpoint checks.

A future dynamics reconstruction additionally needs thrust vectors and throttle history, mass depletion/staging, attitude and power constraints, solar radiation pressure and third-body forces, navigation maneuvers, asteroid gravity/shape, and validated initial states. Subsystem matching further needs actual mode loads, thermal node properties, efficiencies/degradation, ground contact schedules, RF losses/coding and reserves. The sources above do not supply a complete correlated flight model or covariance. Missing inputs should remain unavailable or explicit sensitivity assumptions, never silently promoted to mission facts.

## Final packaged-data review

The final packed references were checked with [the independent review script](review_imported_references.py); [the full results](import-independent-review.json) include source SHA256 values and equality checks against the reference embedded in each downloadable LOTUS analysis snapshot. Source and embedded reference values match exactly. Public report/data/source copies match their research originals byte for byte, and all 21 retained source-file hashes match the manifest. Research provenance remains dated **10 November 2026**.

| Packed reference | Usable coverage ends, UTC | Independent event checks | Largest event position / velocity error | Held-out interpolation checks | Largest measured interpolation position / velocity error | Excluded intervals |
|---|---|---:|---|---:|---|---:|
| BepiColombo | 20 Nov 2026 00:00:00 | 9 | 0.695 m / 0.02041 m/s | 36620 | 100.133 m / 0.09322 m/s | 59 |
| Psyche | 1 Mar 2027 11:59:59 | 1 | 0.580 m / 0.000046 m/s | 8464 | 99.852 m / 0.08774 m/s | 17 |

Event checks compare imported spacecraft-minus-planet vectors with the independently requested direct planet-centered vectors at the **same nearest-second UTC instant**, after the declared frame rotation. Interpolation statistics describe tested held-out states outside excluded source intervals; they are not covariance, a guaranteed global error bound, or a statement of actual spacecraft navigation accuracy. Packing slightly changes the BepiColombo measured maximum from its prepacking value, so the final measured **100.133 m** figure is retained explicitly.
