Robert's Lambert Multi Gravity Assist Solver
Explore the live trajectory gallery · Desktop downloads & source · Robert’s portfolio
The live GitHub Pages explorer combines curated mission examples with a shared trajectory library. New shared submissions join the destination filters, playback, mission dates and downloads automatically. Duplicate same-day encounter sequences share one public entry.
Curated Lambert admission independently recomputes flyby speeds and turning geometry: excess-speed mismatch ≤1 m/s, each DSM ≤700 m/s, and inferred periapses above their stated floors. Coast endpoints and event states are cross-checked against the independent JPL cache. Integrated concepts have a separate Sun/eight-planet clearance audit. See docs/data/catalog_audit.json.
Shared pass/yellow labels preserve the original submitting limits and are not independent validation. The desktop app recomputes their science before reuse.
The website is static HTML/CSS/JavaScript under docs/. It runs no optimizer and requires no Python server, account or API key. Interactive playback interpolates saved points. The desktop application below performs new searches on the user’s hardware.
Explore interplanetary gravity-assist sequences, inspect candidates as they are found, refine nearby encounter dates, and export propagated trajectories for Celestia or other tools. All optimization runs on your own CPU; no AI service, API key or subscription is required.
Download the v0.2.5 package for your platform: Windows ZIP, Apple Silicon or Intel Mac DMG, or Debian/Ubuntu x64 DEB. They open the same local browser GUI. On Windows, extract the ZIP and open RoLaMGAS.exe inside the folder. The Mac preview is unsigned/not notarized. See platform instructions. Python and the scientific dependencies are bundled. The eight-planet baseline is included, covering 2027–2100 at one-day cadence, with extra six-hour Mercury data. The first run installs those tables into your persistent cache.
Searches, exports and extra Horizons downloads are stored in
%LOCALAPPDATA%/RoLaMGAS, separate from the executable. Updating the executable
preserves your work. An internet connection is needed for targets or epochs not
already cached. ROLAMGAS_DATA_DIR can select another data directory.
Finite flyby exports now jointly refine departure, encounter geometry and DSMs, retain the correct heliocentric winding through resonant returns, and allow intermediate date adjustments up to 365.25 days. All remaining corrections are listed as actual burns. Launch and final encounter dates remain fixed.
The approved CLAUDIUS refinement cuts extra patch corrections from 177.5 to 7.5 m/s and total cruise impulses from 915.1 to 628.4 m/s. The website has 117 validated replacement downloads, with original files, exported dates and burn tables retained. Full-precision website handoffs preserve exact Julian dates. See the release notes.
Proposal initialization now tries every matching historical/study reference and gives each retained reference its own feasibility refinement. In-range reference departures keep their timing instead of being forced onto a C3 boundary. New cited Decadal Survey presets expand route coverage. These are initial guesses, not a guarantee of a feasible mission in a different decade.
The release also fixes hidden-field validation blocking Start Search, loads exact website candidates for local revalidation, and offers optional JPL dwarf-planet patches when exporting Celestia packages. Named finite-radius study downloads include clearance, interpolation and burn reports. See the release notes.
Complete EdvEVEEJ starters now include a solver-built Jupiter route at C3 6.25 with a 493.9 m/s first-leg DSM, plus four higher-C3/lower-DSM tradeoffs. Select Juice style low C3 to load a compatible example. Every starter is rechecked against the current limits; its reference launch is November 2033 and its last Earth flyby is about 204.5 km high. This is an original Sun-only design, not a reconstruction or an n-body validation of JUICE. See seed details.
The results C3/time graph starts folded and supports filtering and candidate selection. Planet markers and paths have separate preview switches. Selected candidates get render priority; red trajectories skip rendering. Combined unresolved flyby speed mismatches count against the DSM screening budget. Efficiency ratings are removed; destination groups remain. See preview details.
Search the shared trajectory database before allocating optimizer workers. Every compatible candidate is revalidated against the current ephemerides and all your search constraints. Zero, one or two verified matches start local optimization automatically. Three or more show a choice: use the database results, or search locally too. Bundled records work offline; service failures fall back to the local solver.
The app includes 2,047 sanitized passing trajectories, connects to the live community API by default, and defaults to showing yellow candidates while hiding red candidates. Sharing has a visible opt-out control. The public library excludes C3 above 250 km²/s² and small-body flights of 60 years or more; local search limits remain configurable. See community data details.
Use 64-bit Python 3.12. On Windows, run Setup Python.cmd, then
Launch RoLaMGAS.cmd. On macOS/Linux, create a virtual environment, install
requirements.txt, and run python rolamgas.py.
The baseline archive is unpacked into cache/ at first launch. New custom-body
data are downloaded from JPL Horizons and cached locally; searches then work
offline within that coverage. The source app saves searches next to its code.
Open a candidate and click I like this one. Choose an encounter-date variation (default ±30 days). A separate search retains its assist sequence, starts from its date/DSM solution and Lambert branches, and minimizes C3 under the saved mission limits. The parent search is retained. A local search cannot guarantee an improvement or a global optimum.
Open its inspector and choose Add a flyby afterwards. Select a planet, number of added encounters and timing bounds. The earlier cruise, dates and launch C3 stay fixed; only the new portion is searched. Optional new-leg DSMs retain the saved burn limits. Start with one Mercury return and extend a result again if its resonance is narrow. Yellow results can still be useful when the requested arrival speed has not been reached.
Choose repeated Mercury encounters, for example EVMeMeMe or
EVVMeMeMeMe. Me means Mercury; M means Mars. Resonance-informed guesses
favor returns separated by integer Mercury years while testing multi-revolution
Lambert branches. Every result is still checked for speed matching and turn
feasibility above the periapsis floor. Minimize arrival V∞ is useful when
seeking a less expensive terminal capture. Capture burns are not included in
the interplanetary screening budget. These templates do not reproduce
BepiColombo's continuous electric propulsion.
JPL geometric heliocentric Ecliptic J2000 vectors, TDB dates, km and km/s. The search uses solar Lambert/Kepler arcs and asymptotic flyby joins. Optional DSMs and powered flybys have explicit budgets. A passing screen is an approximate patched-conic candidate, not a navigation-grade n-body solution. Finite-radius trajectory validation/export reports patch correction burns and clearance checks separately. CSV, XYZV and complete Celestia packages are available.
See GUI_GUIDE.md, SHARING.md and the original JPL response metadata for details.
python -m unittest discover -p "test_*.py"
Install requirements-build.txt and run
python build_release.py --native --version 0.2.5 on the target OS to build
the native EXE, DMG or DEB plus matching GPL source. GitHub Actions can build all
three platforms; see SHARING.md.
RoLaMGAS is distributed under GNU GPL version 3 (GPL-3.0-only); see LICENSE.
You are free to make derivative works based on this code, or improve its functionality.
Third-party libraries retain their own copyrights and licenses; see THIRD_PARTY_NOTICES.md. The source release includes the code, build instructions, pinned dependencies, model assets and baseline ephemerides corresponding to the Windows preview. JPL supplies the ephemerides and does not endorse this application.
GitHub Pages serves docs/ from main, with no build step. The catalog contains
curated numerical results, not user histories or arbitrary cache contents.
tools/build_public_catalog.py accepts a private trajectory project and app-data
directory as inputs to regenerate the public field subset. The existing gallery
runs entirely from the committed static data and does not need those private inputs.