How it works

From source to decision

What it can do

• One operating picture

  Radar, video, AIS, drones and perimeter sensors as one view, not separate screens.

• Event correlation

  Signals from different sources become one incident.

• History that stays

  Nothing is discarded when an alert is closed.

• Three learning loops

  The site profile, Support Center feedback, and de-identified patterns across the network of sites.

• Scenarios in time

  Options are compared before a decision has to be made in a hurry — including a change of route, an unknown contact, or loss of a link.

• Alarms with a next step

  An alarm arrives with a recommended action.

• Reports from history

  Incident timelines, periodic statistics, and post-analysis drafts — checked by a person before they go out.

• No single point of failure

  If one sensor, component or link fails, watch continues in a simpler mode, and the operator is told.

Four layers

• Adapter Layer

  Transition interfaces to equipment and software already on the site.

• Unified Data Layer

  One normalized model: Object, Event, Sensor, Alert, Case — independent of maker.

• Intelligence Layer

  AI, Risk Engine and Scenario Engine — only after the data layer is live.

• Human Layer

  Support Center. The person remains in the loop.

Four levels of response

Not all automation is the same. The boundary does not move.

• Monitoring

  Continuous collection, correlation and status — automatic. The person watches. No action is required.

• Assisted Analysis

  AI forms a risk assessment, scenarios and recommendations. The person reviews and decides.

• Supervised Automation

  A pre-agreed sequence runs: alerts, recording, access restriction, escalation to Support Center. The person is notified and can stop it.

• Manual Override

  The system carries out a direct instruction from the responsible person immediately, ahead of the queue.

Supervised Automation is limited to reversible procedural actions — alerts, logging, watch systems, access restriction under protocol. Physical countermeasures are taken only by a person, through licensed equipment, where the law allows.

Explainable risk

A defined scale — not an opaque score. The reasons behind each change stay visible.

• NORMAL

  Ordinary conditions. No deviation.

• ATTENTION

  A change that should be watched.

• ELEVATED

  Several factors reinforce each other.

• HIGH

  Active attention from the responsible person is required.

• CRITICAL

  Crisis mode and priority accompaniment by Support Center.

Alarms in seconds

The first notice is sent at once. Context and a recommended next step follow as analysis completes. The captain does not wait for a full report to learn that something happened.

• Detect

  A sensor, or several sources together, marks a deviation.

• Classify

  A preliminary risk level and event type are assigned.

• Notify

  The responsible person — and Support Center when needed — are reached without a further delay for checking.

• Recommend

  A concrete next step is offered, not only a fact.

• Protocol

  Where a situation was described in advance, Supervised Automation can start.

Two ways to decide

• Known crises

  For situations described in advance — a person in the water, collision risk, a perimeter breach, fire — StarWall walks the crew through a fixed sequence. No improvisation under stress.

• Atypical situations

  If the situation does not fit a protocol, StarWall ranks options and their risks and brings in a Support Center specialist on the same picture the captain sees.

Five rules that do not wait for a later version

• No single point of failure

  Loss of a link, a sensor or a temporary AI module must not leave the site without basic watch.

• Degrade, do not stop

  With partial data the system continues in a simpler mode and says so.

• A person in the decision

  No automatic conclusion replaces the responsible person in a critical situation.

• Explainable alerts

  Each alert can be shown: which data and which rules produced it.

• Privacy by default

  Site data is not used to train the network in an identifiable form.

A copilot, not a panel

StarWall is a continuously working digital copilot for a yacht captain, or for the person responsible for security at a marina, a port or an island. It works on three foundations at once.

• Calculated information

  The profile and history of the site.

• Live data

  What the system sees right now.

• Direct instruction

  The instruction of the responsible person — taken first.