...
- MAPGEN as a model for HYPSO is important, since we will be using the 'mixed-initiative' (human-in-the-loop) control methodology. See GDS slides at: <GDS slides on this Wiki> to Documentation to show the recommended evolution
- Fundamentally there is a front end visualization framework (JPL's APGEN software) and the back end AI Planner
- We cannot use MAPGEN since it is export controlled and old nowdated
- Open-Spify is the "new" MAPGEN with caveats
- It can come with and with a new Java front-end + constraint engine + planner (optional)
- The constraint engine does local propagation to detect planning/scheduling conflicts but does not resolve the conflicts. With a planner, the resolution occurs aided by a model with heuristics to minimize search.
- There can be two types of planning engines (both EUROPA based https://github.com/nasa/europa/wiki/Europa-Background)
- Batch and dynamic – typically batch processing implies a set of goals are fed into the planner, when the crank is turned the planner generates a possible plan, as an outcome (assuming a plan is viable). This was what MAPGEN was doing on MER.
- A dynamic planner will continuously modify an output plan while interacting with the front end and/or other sources of information.
- For HYPSO, we will rely on the simpler form of (static or batch) planning and push to look towards how to formulate resource constrained plans.
- This will ensure, we are building something useful operationally for HYPSO, and then build some novel resource optimization work on top ideal for a PhD dissertation.
- More info on Spify being collated.
...