03 / Autonomous Systems & Robotics
MissionGuard
Runtime Assurance for Fleet-Level Mission Autonomy
A fleet can fail a mission while every vehicle in it looks healthy.
Mission autonomy increasingly behaves like a black box: a planner emits a course of action, and operators are asked to trust it. Per-vehicle health monitoring does not answer the question that matters, which is whether the plan itself still satisfies the mission.
MissionGuard never inspects the planner. It adjudicates each emitted course of action against five declared mission contracts: resource feasibility, coverage persistence, task deconfliction, contingency response, and information freshness. The contracts are runtime-checkable claims about the plan, evaluated continuously as the world changes.
When a contract is violated, MissionGuard raises a fault card that names the violated contract, carries the evidence for the determination, and recommends a specific mitigation hook, such as forcing a replan with the active contingency injected as a hard constraint, or rejecting a stale COA keyed to an outdated world version.
Every alert is auditable end to end. The prototype scores itself with randomized scenario sweeps, reporting precision, recall, time-to-detect, and traceability coverage, and exports the full fault log as JSON and CSV evidence.
The assurance loop
- Mission intent Tasks, regions, coverage requirements, and priorities are declared up front.
- Autonomous fleet A black-box planner assigns simulated platforms to tasks and emits a course of action.
- Runtime observation The assurance layer watches world state: platform status, threats, comms, and information freshness.
- Contract validation Each COA is adjudicated against the five declared mission contracts.
- Mission assurance Violations become fault cards with evidence and mitigation hooks, exported as auditable logs.
Inside the system
Working on problems MissionGuard addresses?
We collaborate with government organizations, research institutions, and technology partners. Tell us what you are trying to solve.