Orbit lets a planner inspect a violation, pick a lettered fix or Waive, or send the set to AI — then confirm without giving automation the final word.
Teague / UCI capstone concept · Not a NASA product
SME interview plus published SPIFe, Playbook, HERA, and CAST evidence framed the problem.
Stage 4 and 5 of the FAO journey: detect conflicts while building, then resolve them.
A working Figma-plus-code prototype instead of static mocks.
At 08:00, two activities need the same treadmill. Somewhere in the week a rule has broken, and a Flight Activities Officer has to find it, understand it, and choose a repair.
Playbook and SPIFe-era tools already flag and describe violations.5 But the rule, the affected activities, and the repair action live in separate places. Detection without a decision surface.
23.8%1
1,617 of 6,802 flexible activities rescheduled across four 45-day HERA C6 missions. All crew used Playbook for spontaneous edits.
Dozens of people, multiple days2
A week-long ISS crew plan is built far ahead of execution. Re-planning under new constraints is similarly taxing.
First ISS self-schedule, 2016-173
CAST: five exercises, Dec 2016-Jul 2017. First time an ISS astronaut scheduled their own tasks and executed as planned. Tech demo, not ISS ops of record.
Flight Activities Officers and operations planners building and repairing a mission week.
Keep the violated rule, affected activities, lettered repairs, Waive, and confirm in one decision context.
Capstone concept scoped to one flow: detect a conflict while building, then resolve it in schedule context.
In HERA C6, most spontaneous edits were small shifts, not full replans.1 The repair space is full of small, comparable moves that a system can propose and a planner can compare.
The matrix plots updating activities, recovery, logistics, monitoring systems, scheduling, readiness, reentry monitoring, crew communications, navigation, telemetry, payloads, simulation training, diagnostics, workarounds, Go or No-Go calls, abort decisions, and emergency procedures. Scheduling appears at guarded automation risk and lower implementation effort; abort decisions and emergency procedures appear at higher risk and effort.
Team Rocket's internal synthesis recorded a firm guardrail: reserve AI for problems ordinary automation cannot solve, and stay inside one specific use case.6
Internal synthesis of a scheduling-SME interview · Jul 2026
Team Rocket combined SME interviews, secondary research, and comparative analysis across SPIFe, Playbook, HERA, CAST, and adjacent mission systems. For this prototype, the team used an internal scheduling-SME synthesis to evaluate six directions against two filters: choose a genuinely hard problem for AI, and stay inside one specific use case.6
Constraint visualization
Keep the violated constraint and affected activities inspectable in the schedule context.
AI-assisted violation resolution
A large, non-obvious repair space where automation suggests and the planner makes the final call.
Constraint creation
Promising, but upstream from the violation moment and therefore a second use case.
Resource trade-off dialogue
Useful, but conversational AI plus resource modeling would widen the prototype beyond one flow.
Resource matching
High value, but ordinary search and filtering can solve it without generative AI.
Historical pattern detection
The necessary schedule-change dataset was not available to the team.
Baseline systems already explain violations.
HERA showed no-go zones, violation descriptions, and suggested fixes.5
Co-locate explanation with action. Orbit does not reinvent detection; it fuses the rule, activities, and repair into one surface.
The planner keeps the last word.
“The human is in control of the plan, but he or she may selectively invoke help from automated systems.”4
McCurdy et al., ICAPS 2011
AI proposes, never applies. Continue to confirm commits. Start over and Waive stay visible.
The full eight-stage FAO journey. Stage 4 and 5, detect and resolve, are the prototype scope.
An auto-repairing scheduler would be faster; it would also change a mission plan silently.
The system fixes violations as they appear. Faster, but the plan changes without the planner, and decades of mission-ops precedent keep the human in control of the plan.4
Every repair costs one more click. In exchange, the schedule never changes without a deliberate confirm, and Waive keeps relaxed constraints visible instead of buried.
Orbit trades automation speed for planner authority. That is a trade worth making.
SPIFe research calls for logged waiver rationale. Capturing that rationale at confirm is the next design step.
Sidebar → top nav
The persistent mission tree competed with the task at hand. Removing it lets the app open as a focused page instead of a dense workspace.
Conversation → instrument
A chat transcript hid the schedule it was changing. Steps, source citations, constraint tabs, and a live week preview put the reasoning next to the plan.
Partner → engine
Open-ended chat gave the AI a vague job. A violation list, then a resolution rail, narrowed it to one: propose repairs the planner can compare.
HERA and CAST document crew self-scheduling and small schedule shifts; SPIFe documents human-in-control planning.1-5
A capstone concept, not a NASA product; no usability study or operational metrics yet.
Ask FAOs to identify the broken constraint, compare lettered fixes against Waive, and judge whether the AI receipt supports the choice before confirm.
Designing AI meant defining where it should stop
Locking the story to one person and one moment. FAO, constraint violation, last word with the planner.
I would test constraint hatching and the resolve panel before expanding the surrounding product surfaces.
Use AI on the hard problem. If ordinary automation can do it, do not call it AI.
- Abbott, Karasinski, Marquez, Characterizing spontaneous self-scheduling in NASA's HERA campaign 6, IEEE Aerospace 2025.
- Hillenius, Marquez, Korth, Rosenbaum, Evaluation of crew onboard planning: year 2, NTRS 20180000770.
- Marquez, Hillenius, Healy, CAST slides, NTRS 20180005211.
- McCurdy et al., SPIFe / ICAPS 2011.
- Marquez, Shelat, Karasinski, HERA C6 slides, NTRS 20220013438.
- Team Rocket, internal synthesis of a scheduling-SME interview, Jul 2026.
More adventures in design
