Prev-Talk | Next-Talk | All-Talks | Talks-Sorted | MOOS-DAWG'26
Talk-33: Talking to MOOSDB: An LLM Experiment for Non-Developers
Ian Cheng
JetSea AI, Taiwan
Current MOOS development tools are excellent for developers. They are what we use every day. But as we delivered MOOS packages to customers this past year, we found that an interface built for developers can overwhelm people who are not developers. The users we now build for split into two groups: customers, everyday operators who need a lower barrier to entry, and commanders, who need a command-and-control display to make quick decisions.
We think LLMs might be part of how MOOS-IvP reaches people who aren't developers at all. Instead of clicking through menus or learning which button does what, an operator can simply talk to MOOSDB directly? describe what they want, and let an agent figure out how to carry it out. This talk treats that as an open experiment: what changes when an LLM sits between the operator and the MOOSDB? We walk that up from substituting a single command, to standing watch over a fleet, to being handed an intent and asked to work out the rest.
We'll share an early prototype built around this idea, a lightweight dashboard paired with an AI chat box, including what we've learned so far, where we deliberately keep a human in the loop before anything happens on the water, and a short live demo. It's still early work, and some of what we learn may be worth bringing back into tools like pMarineViewer itself. We hope to spark discussion with the community.
Categories:
- Mission Control
- LLMs