what robot fleet monitoring means :
Robot fleet monitoring is the continuous collection of status, health and performance data from every robot in a fleet, shown in one place, with alerts when something needs attention.
it is the observing half of robot fleet management. management acts on the fleet. monitoring is how it knows what to act on.
three layers to monitor :
- hardware health : battery, motors, temperatures, sensors, connectivity. is the machine able to work.
- software state : which processes are running, the rate of each sensor topic, error states, and the version of everything on each robot. a camera topic at 0 Hz is a software fact before it is a task failure.
- task performance : missions completed, time per task, interventions, and whether the task itself succeeded. is the work getting done.
most dashboards are strong on the first layer, decent on the second and thin on the third, because the third needs a definition of success that the robot cannot supply on its own.
the metrics worth a dashboard :
- availability : the share of time each robot was ready to work
- throughput : tasks or missions completed per hour
- intervention rate : how often a person had to step in
- time to recover : from a stop to the robot working again
- task success rate : per task and per operating condition
- version drift : robots that should match and do not
the first four are operations metrics. the last two are reliability metrics, and they are the ones that explain a bad week. how to state a success rate with its interval is in measure robot task success rates.
alert on trends, not only on states :
a state alert fires when something is wrong now : a robot is stuck, a battery is low, a mission is late. it is the right alert for the person running the shift.
a trend alert fires when something is drifting : the success rate on one task is down against its baseline, one robot needs more interventions than its twins, a sensor rate has been sagging since the last update. it needs a baseline to compare against, which is why few fleets have one.
the blind spots :
- uptime is not success. "97% uptime" says the robots were available. it does not say the pick succeeded 97 times out of 100 under the lighting on the night shift.
- the robot reports on itself. a mission marked complete is not always a task that succeeded, and when a sensor drops out the log goes quiet exactly where the answer was.
- averages hide units. uneven motor wear or a firmware version that differs on three robots does not show in a fleet-wide number.
- the conditions are missing. the lighting, the layout and what was left in the aisle are outside the robot's own telemetry.
an independent record of the conditions around each run covers the last two. that is the role of a robot fleet supervisor.
from an alert to a cause :
an alert starts an investigation. it does not finish one. keep the recording of the run that triggered it, find a healthy run of the same task, and compare the two with the code, model, hardware and conditions behind each. the workflow is in investigate a robot field failure and compare MCAP robot runs against a baseline.
when the cause is an update, the fix is to catch the next one before it reaches the fleet. see robot regression testing.
where OORB fits :
OORB is not a fleet monitoring dashboard, and it does not replace one. keep the platform that shows you the fleet right now. RunFP adds the record behind the numbers : every run with its code, model, hardware and conditions, a reliability figure per task and condition, and a ranked list of what changed when results drop.
we compare the two workflows, with the vendors' own documentation, in OORB RunFP vs Formant and InOrbit.
frequently asked questions :
What is robot fleet monitoring ?
Robot fleet monitoring is the continuous collection of status, health and performance data from every robot in a fleet, shown in one place, with alerts when something needs attention.
What should you monitor on a robot fleet ?
Three layers: hardware health such as battery, motors, temperatures and sensors; software state such as running processes, message rates, errors and versions; and task performance such as completed missions, interventions and success rate per condition.
What is the difference between fleet monitoring and fleet management ?
Monitoring observes: it collects data and raises alerts. Management acts: it assigns tasks, coordinates traffic and rolls out updates. Most fleet platforms do both, and monitoring is the part that tells the rest what is happening.
Does fleet monitoring show whether a robot is reliable ?
Only partly. Monitoring shows that robots are available and that missions were reported complete. Reliability needs a defined task outcome, a success rate per operating condition and a baseline to compare against.
the dashboard shows what the fleet is doing. the record shows whether it is doing it well.
bring the run data behind one alert you could not explain and we will show you what was different about it.
Robot failing in the field? Get help now