Aplark AI

A spacecraft reports hundreds of telemetry channels. Fixed limits catch large faults but miss slow drift or unusual combinations of channels. Reviewing every channel by hand does not scale as the fleet grows.

Aplark AI is designed to provide the following.

• Baselines of normal behaviour learned from telemetry history.
• Detection of deviations in single channels and across channels.
• Findings ranked by how far they depart from the baseline.
• Evidence shown with each finding: channels, time range and comparison period.
• Operator feedback recorded so false alarms can be reviewed.

An example of how an operations team would use Aplark AI. The steps are illustrative.

1. Telemetry history is loaded for a spacecraft.
2. A baseline of normal behaviour is built and reviewed with the engineers who know the spacecraft.
3. New telemetry is compared against the baseline.
4. A finding is raised with its evidence, and sent to the operator in MissionOS.
5. The operator marks the finding as confirmed, explained or false alarm.

Aplark AI reads telemetry through MissionOS and sends findings back to it. It needs telemetry history from the partner spacecraft. The amount of history and the channel list are agreed with the partner.

• A baseline for a defined set of telemetry channels.
• A review of findings on the partner historical data, with the engineering team.
• Measured false alarm and detection rates on that data, with the method used.
• A written note on which channels are suitable for this approach.

Pilots are scoped with each partner. See the Design Partner Program.

Aplark AI does not send commands. Findings depend on the quality and length of the telemetry history. Detection and false alarm rates are measured per spacecraft, not assumed.