Topic E — Feature opportunities
Follow research-brief.md for the output format and
rules. Save the result as
research/findings/E-features.md.
Run this topic after A to D and read their findings in
research/findings/ first. It depends on what competitors
lack (A), what the data supports (B), what users need (C) and what
technology exists (D). If a finding is not available yet, do the minimum
checks yourself and say so.
- Feeds: 1.1 Technical Objective; 1.2 Requirements; 1.4.4 Implementation aspects; 1.7 Technical Programme of Work (candidate work packages).
- Find: capabilities, within or beyond the current SELIS description, that would make the proposal stronger. Look at what comparable tools offer, what users and ESA roadmaps ask for, and what recent published methods make possible.
- Innovation is welcome. A feature may go beyond what the raw data shows directly, and may use a method nobody has yet applied to lunar simulation, as long as each step is grounded in something that exists and can be cited. Ideas with no such grounding are rejected.
The five gates
Every candidate feature must pass all five, with evidence, or it is rejected.
- Data basis exists: the feature rests on data available now for the lunar South Pole under terms that allow this use. Name the data set. The data may be augmented (super-resolution, gap filling, approximation, ML, algorithms that extract signals not visible in the raw product), provided the augmentation follows a known pipeline, code base, algorithm or documented procedure. Name it. A feature with no real data underneath it fails.
- Method is grounded: either a published method or working implementation for the lunar case, or a method proven in another domain (Earth observation, Mars, terrestrial robotics, computer vision), or at minimum a research paper reporting promising results in that direction. Name the source and say what would have to be adapted for the Moon. A feature that depends on an unpublished or purely hypothetical technique fails.
- Deliverable in scope: a small team could deliver and demonstrate it within a Type A activity, as a proof of concept at TRL 3. Give a rough effort class (weeks, months, more than six months) and the main dependency. A feature that needs an integrated, validated system to show its value fails.
- Verification path is grounded: there is a credible way to show it works, based on how prior research or feasibility studies validated the same or a similar approach (comparison with real imagery, held-out measurements, a reference data set, a published benchmark). Cite the precedent and state what the check would be here.
- Wanted: a named user group or an ESA programme has a stated need for it. Cite the evidence.
Evidence grade
Grade the evidence for gates 1, 2 and 4, taking the weakest of the three as the feature's grade.
- A — proven for the Moon: demonstrated on lunar data.
- B — proven elsewhere: demonstrated in another domain; transfer to the Moon is plausible and the adaptation is described.
- C — research-stage: supported by published promising results only.
For grade B and C features, state the main technical risk and a fallback (what is delivered if the approach underperforms).
Output specifics
- Table columns: feature, user need and evidence, data basis and augmentation, method and its source, evidence grade (A, B or C), effort class, verification path, main risk and fallback, verdict.
- Verdict is one of: core (belongs in the Type A scope), option (credible add-on or later phase), reject (state which gate failed).
- Grade C features should normally be an option rather than core, so the core scope does not depend on unproven work. If one is proposed as core, justify it and give the fallback.
- Keep the rejected list. Knowing why an attractive idea fails is useful, and it stops the idea returning later.
- Aim for at most eight candidates passing, ranked by value to the proposal against effort. Do not pad the list.
- Stay honest about the four data classes in
CLAUDE.md: a feature that produces inferred or synthetic output must say so and say how it is labelled. - Answer: which set of features forms the strongest Type A scope, and which two or three additions most improve the proposal for the least delivery risk.