This article provides general information. Any real project must be assessed against its site, intended use and the rules in force at the time of deployment.
Define a verifiable task before the route
A successful patrol must provide a useful observation, transmit it and allow discrepancies to be handled. Simply finishing the route does not demonstrate these three results. For each point, describe the question to resolve and the person who will use the information.
For a car park, context includes authorised movements and vehicles that may obscure an area. For a warehouse, it also includes closing handover and late activities. The task must be reassessed when these uses change.
Repetitive patrols are the first practical use case
Professional sites often contain secondary entrances, doors, fences, loading bays and outdoor stock. A robotic patrol can visit those points on a schedule, pause at defined positions and report anomalies.
Robots are most valuable where the route is stable, missions are frequent and the information being sought is clearly defined.
What the robot can check
Depending on its sensors, it can provide images, identify a presence, measure apparent temperature, indicate that a door appears open or confirm that a route is clear.
- Fences and gates.
- Loading bays and delivery areas.
- Car parks and outdoor yards.
- Sensitive stock or isolated equipment.
- Technical checkpoints accessible from the route.
Mobile observation to support alert verification
When a fixed sensor raises an alert, mobile observation may offer another angle if the terrain, configuration and integration allow it. The pilot must check actual access time, information quality and what happens when the robot is unavailable. No time saving should be assumed.
Reduce unnecessary call-outs
Animals, moved objects, weather and reflections generate alerts. Mobile verification can avoid deploying an officer unnecessarily while preserving a useful event record.
A robot does not replace every human security task
Robots do not handle every social situation, conflict, emergency or unexpected obstacle well. They also lack the judgement and intervention capabilities of a trained professional.
Assign the roles
The machine completes repetitive missions and gathers information. An operator interprets ambiguous situations, decides escalation and coordinates responders. This division is more realistic than complete replacement.
Avoid monitoring employees
In Switzerland, a system cannot be used to monitor employee behaviour. Patrols justified by security should minimise the presence of employees within the camera’s field of view.
Programme missions around time and place
Night-time missions on a closed site are easier to justify than continuous monitoring of a workplace. Employees need advance information, unmonitored areas should be defined and access to footage restricted. For verification, see Video surveillance in the workplace (FDPIC).
The robot’s function should appear in internal documentation. Data collected for security must not be repurposed for individual performance evaluation.
Measure the value of a pilot
Return on investment is not simply the number of patrol hours. Track completed missions, detected incidents, avoided false call-outs, downtime and time spent on manual recovery.
Begin with a limited route
A first pilot can cover a small number of high-value checkpoints during a defined period. This tests connectivity, terrain, weather, acceptance and monitoring integration before the system expands.
A pilot protocol that also exposes limitations
Choose announced tests with participants, without putting anyone at risk. Record the scenario, conditions, expected result and actual result. Acceptance thresholds must be agreed before the trial; we propose no universal success percentage.
The following table is an original preparation tool. It does not report the performance of a tested robot. A condition not encountered during the pilot must remain marked as unvalidated.
| Agreed test | Expected evidence | Discrepancy to document |
|---|---|---|
| Normal task on the selected route | Points observed and useful information transmitted to the recipient. | Point reached but observation unusable. |
| Authorised test passage through a zone | Detection and context available according to the task. | Missed event or unnecessary notification. |
| Temporarily inaccessible checkpoint | Incomplete task reported and fallback procedure known. | Patrol presented as complete despite the missing checkpoint. |
| Connection fault tested safely | Known state, expected behaviour and fault sent through an available route. | Lost information or no fault report. |
| Restart after correction | Return to a checked configuration and repeat test. | Setting changed without checking the original task. |
Measure the work actually required
Record scheduled missions, those that provided all useful observations, missed test events, alerts requiring no action and manual recoveries. Distinguish travel time, time to make an image available and human decision time.
Add cleaning, updates, troubleshooting and route preparation. Fewer human patrols do not prove a saving if other tasks shift to operations. Compare documented expenditure and tasks actually covered before deciding to expand.
Frequently asked questions
Can a robot replace security-officer patrols?
It can automate selected repetitive patrols, but decisions, complex situations and sensitive interventions remain the responsibility of trained people.
Can a robot operate during working hours?
Potentially, but safe movement and the prohibition on monitoring employee behaviour must be built into the project.
Which sites are most suitable?
Sites with stable routes, large areas, repetitive missions and clearly defined checkpoints generally offer the best conditions.