Designing Dispatch Screens for Robotics


Collaborators

  • William Martin Engineering & Design
  • Karen Kosasi UI/UX

How I approached designing dispatch screens for a fleet my first time

I had to view the problem as multi-layered. There are 4 constant rules. People set the rules, watch the outcomes and handle what the rules can’t.

Facility Fleet Manager All Systems Nominal 10:24 AM

Notifications

    Total Robots5
    Active5
    In Queue1
    Alerts0
    Supply Stocker(SS-01) Restocking Lobby Stockroom 62%
    Supply Assistant(SA-01) Fetching Battery Pack 43%
    Window Washer – WW-01 Active 10:24 AM

    Current Task

    Wash East Tower Windows

    71%

    Est. Completion4 min

    Operation Details

    • Apply cleaning solution
    • Scrub and rinse
    • Dry and inspect

    Building Zones

    • Complete
    • In Progress
    • Pending
    Traffic Manager – TM-01 Active 10:24 AM

    Current Task

    Maintain Safe Passage

    All ClearMonitoring Paused

    Work Zone: Lobby North

    Recent Actions

    • Placed Wet Floor Signs
    • Cordoned Work Area
    • Rerouted Foot Traffic

    People Flow

    Clear
    Bathroom Cleaner – BC-01 Active 10:24 AM

    Current Task

    Clean Restroom 2F

    78%

    Est. Completion2 min

    Recent Activity

    • Area disinfected
    • Floors scrubbed
    • Waste tank OK

    Location

    1 / 4

    Every button on these screens works. Use the arrows to move between them, or open one full size.Swipe between the screens, then tap one to open it and try the buttons.

    01

    Route

    Which path each robot takes, and who yields at crossings.

    02

    Assign

    Which robot takes which task, weighing distance, battery, capability and load.

    03

    Sequence

    Which task goes first, based on deadlines like the next truck out.

    04

    Sustain

    Charging and maintenance, so enough robots are ready for the next hour.

    1 / 4

    Swipe or drag the cards to read each rule

    The software built to manage these robotic fleets suffered from a fundamental mismatch.

    Most fleet management tools were built like logistics dashboards for point-to-point delivery trucks (moving a box from Point A to Point B).

    Cleaning and facility robots operate differently in physical space.

    They cover areas, not just routes; they run on tight schedules between shifts; and instead of just running out of battery, they run out of clean water, fill up recovery tanks, tangle brushes, and get blocked in congested zones.

    How cleaning fleets differ from other robot fleets

    Scrubber-dryers that scrub and squeegee, vacuums and sweepers, and trash runners.

    Operating factorsMoving goodsCleaning floors and hauling trash
    01The workA trip from A to BCovering an area, or a full bin to the compactor
    02WhenOn demand, all shiftScheduled, often between shifts or overnight
    03SuccessThroughput against planCoverage, quality and proof it happened
    04What runs outBatteryBattery, clean water, dirty-water tank, bags, brushes
    05Who tends themSupervisors and techsJanitorial or facility staff, often a contractor
    06Typical problemsBlocked, lost, faultsTank full, brush tangled, area busy, zone skipped

    People on the floor

    Shift supervisor

    Key question

    Is the fleet healthy, and where do I need to be?

    Device / channel

    Console and wall display

    Floor associate

    Key question

    Is this robot here for me, and what do I do with it?

    Device / channel

    Robot screen, lights, handheld

    Maintenance tech

    Key question

    Which robot, what’s wrong, and where is it?

    Device / channel

    Handheld or tablet

    Operations manager

    Key question

    Are we on rate for the shift and the week?

    Device / channel

    Desktop analytics

    Vendor support

    Key question

    What happened, and can I see the logs?

    Device / channel

    Remote console

    When things went wrong, human operators (shift supervisors and maintenance techs) were hit with a flood of generic status updates and cascading alarms. The core design problem wasn’t a lack of data; it was too much noise and no clear hierarchy of meaning.

    Diagram of four interlocking loops: two yellow loops running side to side and two white loops running top to bottom, each with an arrowhead showing it circles back on itself.

    With robotics, the floor writes the constraints.

    Office floor plan drawn in white lines: desks, a meeting room, a kitchen counter, restrooms, a lounge with a sofa and armchairs, and the entrance, with a large yellow circle over the upper right of the floor, where the plan is drawn in dark lines, and a small robot driving a delivery loop through the office.

    Night-shift console

    Cleaning robots need coverage views

    Cleaning coverageA–D
    Zone A · cleaned
    Zone B · cleaned

    S-2 · dirty
    tank 85%

    Zone C · cleaned

    Zone D · skipped
    Pallets in
    aisle, 02:10

    • Cleaned
    • Skipped
    • Robot
    1. 01

      Coverage, not dots. Show what got cleaned, what was skipped and why.

    2. 02

      Consumables as countdowns. Dirty-water tank full in ~15 min beats a sudden stop.

    3. 03

      Schedule around people. Clean in quiet hours; pause busy areas, and say so.

    4. 04

      Wet floor is a safety state. Show where it’s still damp; keep walkways clear.

    5. 05

      Proof of clean. A morning report for facility managers and contractors.

    6. 06

      One floor, two fleets. Cleaners yield to dispatch traffic; show both on one map.

    Level 1

    Perceive

    What is happening?

    Example

    R-14 has stopped in aisle C.

    Level 2

    Comprehend

    What does it mean?

    Example

    R-14 is blocking six robots. Station 3 has no work.

    Level 3 · where the value is

    Project

    What happens next?

    Example

    At this rate the shift misses target in about 12 minutes.

    Borrowed from control rooms

    Fundamentals

    • Design for meaning and prediction, not just status
    • Use color as a signal, not a decoration
    • Apply alarm discipline
    • Automation should be self-explanatory
    • Someone should own each decision

    Gray is normal.
    Color means trouble.

    Example · zone overview
    • R-01Moving
    • R-02Station
    • R-03Moving
    • R-04Charging
    • R-05Moving
    • R-06Idle
    • R-07Station
    • R-14Blocked
    • R-09Moving
    • R-10Moving
    • R-11Charging
    • R-12Station

    Shifting the paradigm: the four rules and industrial calm

    To approach this complexity, the design had to pivot from traditional software UI to something inspired by industrial control rooms, specifically high-performance HMI standards like ISA-101 and alarm management guidelines like ISA-18.2.

    The system was rebuilt around four constant operational pillars:

    01

    Route

    Which path each robot takes, and who yields at crossings.

    02

    Assign

    Which robot takes which task, weighing distance, battery, capability and load.

    03

    Sequence

    Which task goes first, based on deadlines like the next truck out.

    04

    Sustain

    Charging and maintenance, so enough robots are ready for the next hour.

    1 / 4

    Swipe or drag the cards to read each rule

    To prevent operator burnout, the interface adopted a philosophy of silence:

    Gray is normal, and color means trouble.

    Muted backgrounds and flat shapes gave the eye nothing to chase when everything was running smoothly.

    Saturated color was strictly reserved for anomalies, always paired with explicit text or shape descriptors rather than relying on color alone.

    Alarms were treated as literal requests for action, capped at a manageable 6 to 12 per operator per hour to eliminate alert fatigue.

    Alarm design principle

    An alarm is a request for action

    If nobody needs to act, it isn’t an alarm.
    Log it as an event.

    Every alarm should arrive with three things: its priority, the likely cause, and the recommended response.

    Watch for floods: one root cause can trigger dozens of alarms at once. Group them, or people start muting everything.

    ~6/hr

    alarms per operator

    Very likely acceptable

    ~12/hr

    alarms per operator

    The maximum manageable

    ISA-18.2 guidance from the process industries. A starting budget, not a robotics rule.

    The view from the robots’ cameras was insightful for understanding the constraints and the environment, as the AI mapped objects and people correctly.

    Fleet manager

    Fleet overview
    Monitor the floor, triage exceptions and route the right person to each robot.ClickTap to view full size

    Promising Results, But a Careful Path Forward

    By designing for meaning, prediction, and alarm discipline, early testing showed a dramatic shift. Operators spent less time chasing phantom alerts and more time proactively managing exceptions. The cognitive load dropped, and human-robot collaboration on the floor felt far more intuitive.

    Yet, bridging the gap between cutting-edge autonomy and messy, unpredictable physical environments is a delicate process. Today, advanced robotics dispatch and telemetry remain very much in beta. Rather than rushing into massive deployments, systems are being scaled slowly and deliberately, site by site, edge case by edge case, ensuring that safety, reliability, and human trust are hardcoded into every deployment before the next wave rolls out.